Every improvement, on the record.

A customer's Timeline can show only your imported data, and its search reaches their whole history

On a customer's record, the Timeline now has All sources, Imported and Website beside the search. Imported shows only the data you bring in from your own systems, so a busy website no longer buries it, and the choice stays set as you move from one customer to the next. Search looks through everything the person has done, in every source: event names, campaigns, channels and every detail of an event.

• When a match is inside an event's details, the line shows which detail matched
• Purchases, Marketing and the other filters show their events straight away, however much website activity came after them
• When nothing matches, it says which filter and which search found nothing

Every opt-out is saved before it is confirmed, and a website event is stored in your warehouse once

An event a browser sends to your warehouse twice is now stored once (Website SDK 1.36.0). The events that name a person, and consent changes, are sent again if a connection drops for a moment, and a later refusal is never outranked by an earlier yes that arrives late. Opt-outs are saved before anyone is told they are done.

• Unsubscribes, including one-click unsubscribes in the person's mail app, and SMS STOP replies are saved before they are confirmed, and a bounce or complaint is accepted only once it is saved
• If an unsubscribe cannot be saved, the page asks the person to try the link again in a moment
• A suppression list import that could not save every address says how many and keeps the list in the import box to import again; addresses already on file are not changed

Send files with a Site chat reply, and one opening line outside your hours

Your team can now send files with a reply on any Site chat, as it already could on an e-mail reply, whether or not the brand lets visitors attach files. Use Attach, drop files on the reply box or paste a screenshot: up to 3 files, 2 MB in all. The visitor sees a photo as a thumbnail and a PDF or document as a named file, each opened through a private link, and the Helpdesk thread shows the files on your message.

• When a visitor who has left gets your reply by e-mail, the e-mail says how many files came with it and that they open in the chat on the site
• Notes, social threads, store reviews and a chat that has moved to e-mail take no files, and the reply box says why
• Outside your Opening hours, Site chat now opens with your Away message instead of the welcome; during your hours it shows the welcome alone, so visitors get one opening line, never both; an empty Away message still shows a short default in the visitor's language

Numbers come in right, however they are written

Every way data comes in – a file upload, a scheduled file, a database or API pull and the webhook – now reads "1 000,50", "1.000,00", "1,000.00", "1 234,50 Kč", "(100)" and "−100" as the numbers they are. How a column writes its numbers is read from the values that come in, and the brand's language settles it only when the values cannot. A value that cannot be read is never counted as 0: it is stored empty, and the import's result says how many there were.

• A refund written −100 or (100) stays negative, IDs, phone numbers, postcodes and dates stay text, and counts such as orders stay whole numbers
• An ad spend row whose spend or a count cannot be read is held and named in the upload's result, and what was stored before for that day stays
• How your numbers are read, on Data › Push, shows what was read for any field whose numbers needed a decision, and owners and admins can correct a column with Change

Sign Out leaves your other devices signed in, and the sign-in forms get a quick security check

Sign Out now signs out only the device you are on, so your other devices stay signed in, and signing in again, with this account or another, opens straight on that account's workspace and brands. When a session ends elsewhere, such as after a password change or a removed seat, the app notices the next time you act or come back to the tab, and the sign-in page says: You were signed out. Sign in again to carry on.

• The sign-in, sign-up and forgotten-password forms run a quick security check that most people never see; it loads only when a form is used, so pages stay fast
• New and changed passwords need at least 10 characters, and the hint says so before you type; existing shorter passwords still sign in

One tag list per brand in Helpdesk, so the same tag is never written three ways

Each brand's Helpdesk now keeps one tag list, started from every tag already in use. Wherever a tag is set – on a conversation, in bulk, in New ticket and in the queue filter – people pick from the list. Case and accents never make a second tag, and close spellings show under Did you mean. Admins and owners get a Tags tab under the queue, beside Reporting, Knowledge and Lookups, to keep the list in order.

• Add, rename, merge, split, archive and restore tags; a tag can be deleted only when nothing uses it
• Suggested tidy-ups finds tags whose names are close, and nothing changes until someone confirms
• Under New tags, choose Anyone on the desk can add or Only admins and owners
• A rename or merge moves every conversation in the background, and each change to the list goes into the audit log
• The list shows 20 tags at a time, with Search the tags, a sort by name or by use, and, once a tag is archived, an Active and Archived switch

When Helpdesk pauses AI on a conversation, the brand's marketing to that person pauses too

When Helpdesk pauses AI on a conversation (Regulated – AI paused), the brand's marketing to that person now pauses as well, until someone on the desk presses Allow AI again: no marketing e-mails, texts, WhatsApp messages, pushes or promos, and no website message meant for that person alone. Service messages and your team's replies still go. The person is also kept out of the audiences you push to Meta and Google Ads, and taken out overnight of audiences already pushed. These conversations are now recognised in every content language the product offers, not only the brand's own.

• The thread says what is paused, beside Allow AI again
• A send's summary counts the person as held – sensitive support conversation, and the push dialog says how many were held; neither ever shows the conversation's subject
• The person is found by e-mail, customer ID or phone, in any letter case
• While the pause lasts, and for people on a Blocks everything blacklist or whose data was erased, no lifecycle stage or purchase prediction is stored and no new loyalty points are earned; points, balances and tiers already earned stay
• Once AI is allowed again, the next push of an audience includes them; an audience already pushed does not take them back on its own

Choose, site by site, what your website script records

Website SDK has a new step after Script installed: What the SDK records. Page views and who the visitor is, conversions and revenue, your own events, forms submitted, consent, site messages, the inbox, Site chat and site tests are always on. Every other capture is a switch for each site, and it stays off – on existing sites too – until an owner or administrator of the brand switches it on.

• Clicks and attention: Clicks, Rage clicks, Buttons seen and Elements viewed; Reading: Scroll depth, Scroll milestones and Page leave; Forms: Forms and Form-field timings; plus Speed and Exit intent
• Each switch says what it gives, which reports need it, an estimate of the analytics events a day it would add, worked out from the site's own page views, and who changed it last
• A report or journey step that needs a capture that is switched off says so and links to Website SDK, instead of showing zeros
• On a touch screen, a rage click now takes 4 taps on the same spot within 1.5 seconds, with nothing on the page changing, so ordinary taps on phones no longer count
• Data shows your Web analytics usage beside your Warehouse events, with the same numbers as Settings › Usage; the warehouse card says whether it counts this month or everything stored in total, and the web analytics card counts the last 30 days
• On Data, the warehouse card's Open the warehouse opens the Warehouse tab

Helpdesk's channel filter works in every view, and an answer to a pending ticket rings

The channel chips in Helpdesk – Email, Chat, Social and By hand – now narrow whichever view is open: Mine, Unassigned, Needs reply, Open, Pending, Resolved or All. With Chat picked, Pending lists only the pending chats, and every count follows. Each chip shows how many tickets in the current view came through it, and clicking it again brings back every channel.

• A customer's answer to a Pending ticket now sounds like any other reply, including one that lands a few seconds after yours
• A chat marked resolved leaves Open at once, as an e-mail already did
• In Site chat, the comment after a rating is now truly optional: Send with nothing written finishes the rating, and the score was saved when it was clicked

A designed e-mail looks like the e-mail, in the Library and in the chat

A designed e-mail saved from the chat now shows as the e-mail itself – its colours, buttons and frame – in the Library panel and on its own page, instead of as a plain article, and the chat's canvas now shows any designed e-mail as the e-mail too. These previews fill in your brand's name, website and logo, and they and the Email Studio canvas show each personal field as a quiet chip instead of raw text such as {{p.first_name|there}}.

• In Email Studio, Sample Data still shows example values in place of the chips
• On the Email Studio canvas, paragraphs are spaced as they will be sent
• An e-mail's unsubscribe spot shows as Unsubscribe link, added when the e-mail is sent, and the Deliverability Score's Unsubscribe link check finds it

Marketing e-mails unsubscribe in your brand's language, and the chat's e-mails keep to what you gave it

The unsubscribe line of every marketing e-mail is now in your brand's language, in every language the product offers a brand – a Czech brand's e-mails say "Nechcete dostávat tyto e-maily? Odhlásit odběr." An e-mail the chat writes carries that line in its own footer's small print, and the unsubscribe link and one-click unsubscribe work as before. Your own brand rules now reach the chat, the AI brand check and the e-mail and SMS writers whole, with the guidance and examples saved with them.

• E-mails written in a language other than English greet in that language, with a fallback word in that language instead of the English "there"
• The chat's e-mails no longer come with guessed open, click or conversion rates or optimisation tips
• The chat's e-mails are now written with only the pictures you, your brand details or a tool gave them: with no picture given, the e-mail is written without one

Save a colleague's e-mail or text in place, in Email Studio and SMS Studio

Anyone who can make changes on a brand can now save their edits, in place, to an e-mail or a text someone else made – a colleague, an AI teammate or the chat. The saved version is the one journeys and sends use. It stays filed under the person who made it, the replaced words are kept in History under the editor's name, and newer changes are never written over.

• In SMS Studio, if someone changed the text after you opened it, Save says so instead of writing over their words, and you can keep yours with Save as new copy under More
• A read-only seat cannot change the saved version: Email Studio keeps its draft and says so, and SMS Studio says to save the text as a new copy under More

Everyone your website names becomes a contact, and their website activity reads in words

Everyone your website names – when your site identifies them, a form submitted with an e-mail, or a conversion that names the person – now becomes a contact of the brand, so the Customers list, its counts and its search find them. No marketing consent is assumed: a new e-mail starts as Not granted.

• A record that already holds their e-mail or your customer ID is completed instead of duplicated, and a value already held is never overwritten
• Two existing records are never merged on their own; Merge stays with a person
• Someone you erased is never added back
• On a site with Require signed identities, only the signed identifier counts
• On a customer's Activity tab, website events say what happened, such as Clicked "Book a demo" on /pricing or Abandoned a form at 40% on /contact, for the captures a site has switched on in Website SDK

A support mailbox connected by IMAP keeps being read

A support mailbox connected by IMAP now keeps being read when the mail server is slow to answer, and mail that arrives meanwhile is never skipped. Only a login the mail server refuses stops it, and that is tried again every 30 minutes. The Support inbox on Integrations says what happened and when, and offers Reconnect or a new password; either brings in the mail that came meanwhile within a minute.

Customers reads like a client book: where you stand, and the one move to make

The Customers page now opens on a headline that says where the brand stands in words – for example how many people are on record, how many of them are buyers, and how many are at risk with what they have spent between them – with one next move beside it, such as winning back the people at risk. That button opens a journey with those people as its audience, and nothing is sent until you review it. Below it come the figures, the stages and the book itself.

• The figures: On record, Buyers, Active · 30 days, Can be e-mailed, Lifetime value and Money in, and, where the brand's data has them, Paid out (or Paid back, for refunds), Value given and Codes used
• Views for Everyone, Most valuable, Recently active, Gone quiet, Can't be reached, New this month and In a conversation, plus your own with Save view, which you can share with the brand's team
• Each row shows the person's stage, lifetime value and where it came from, orders, last order, last activity, how they can be reached and where they came from
• Pick people for Add to an audience, Start a journey, Export or Mark as test; a large export arrives as a file through the notifications bell
• Peek opens a person in a side sheet without leaving the list, and a record opened from the list is read ahead while the pointer rests on its row, so it opens in about a second
• Every figure is read from the brand's own data, by the same rules as the customer record and the personal fields in messages, and Hide test customers keeps test profiles out of the totals
• Counts are exact; one still running reads, for example, "more than 240,000 so far" until it settles

One reading of your money, read from your own data and used everywhere

Every brand now keeps one reading of which of its events are money in, money paid out, value it gives – a bonus, a coupon or a promo – or none of these. It is read from the data itself every night and after each import, with nothing to accept, and owners and admins can correct it under What counts as money on Data › Events. Analytics, the warehouse summary, reports, Flash, Marketing Pulse, lifecycle, loyalty, audiences, the customer record and the personal fields in messages all use that one reading.

• Money paid out and value given never count as money in, and paid out and net show beside money in
• A view across brands never adds two currencies together: each currency is shown on its own
• A test profile's money stays out of every total, while that person's own record still shows it
• The Shopify sync voids cancelled orders and records refunds on their own, so they stay out of revenue

The customer record reads what your own data means, with nothing to set up

The four tiles at the top of a customer's record – Lifetime, the orders, Segment and the last purchase, each in the brand's own words – now come from the brand's own fields, read automatically, and each one says which field it came from or why it is empty. Owners and admins can correct a reading in What your data means, opened from a tile's source line or from the Data manager on Data › Push. It is a correction when you want one, never a step to approve.

• Money shows in the brand's currency, and names stored in capitals show as names
• The record and the personal fields in messages follow one rule, so they never disagree
• Editing a customer keeps every attribute and its type
• A person known only by your customer ID gets a full record – value, activity, movement and stage – with that ID as the heading
• A teammate sees the same money on a record as the owner
• An organisation can switch off the Customer brief, the short AI summary on a record, in Settings › Agents; it is written only when someone asks for it and sends no identifiers

Continue now: move a teammate's next step to today

When a teammate's next step is waiting for a date in its plan, Continue now, in the goal's top bar or on the teammate's row, runs it today. Writing start now or next step to the teammate, in its conversation or by mail, does the same, in English or Czech. Only the current step moves; the steps after it keep their dates.

• A step that waits for your yes, an open question, a paused goal, consent or the budget keeps waiting, and the reply says why
• Asking a teammate for its status says when the next step is planned and how to run it now
• Steps that publish nothing, such as drafting, are no longer spaced out a day at a time

Built for customer bases in the millions, with every person counted

Ad audiences, journeys, privacy requests, lifecycle stages and Home's segments now reach every person they need, however large the brand. A push of an audience to Meta or Google works through every member and carries on by itself, and the audience card shows how many have been checked so far. Before a push, the dialog says how many people go, how many stay home and why; for a very large audience it gives the total and says the counts will appear on the audience when the push is done.

• People who later opt out, lose consent, bounce, are blacklisted, are erased or are marked as test are taken out of a pushed audience one by one; nothing deletes a whole audience on its own
• A removal that cannot be made is tried again overnight, and you are told how many are waiting
• Journeys reach their whole audience, and hand-picked enrolment takes every pick
• A privacy export for a person with a long history arrives in parts, with everything in it
• However large the brand, every customer gets a lifecycle stage, apart from the people the 3 November Helpdesk entry keeps out, and every move fires; Home's segments and the movement matrix count everyone
• Ad audiences, conversion uploads and journey webhooks check each person against your blacklist and their own choices before their details leave the platform

Google, Meta and TikTok Ads Strategists: the measured audit in seconds, and a plan you can apply and undo

Google Ads Strategist, Meta Ads Strategist and TikTok Ads Strategist now share one page design, with Setup, Audit and Your audits. Within seconds of reading the account, the measured report is on screen and saved: a headline worked out from the account's own figures, a strip of five figures, and Where the account stands in plain status words – Fix now, Watch, Clean or Not read – instead of letter grades. The written plan follows, usually within a few minutes. If it fails or you stop it, the measured audit stays, and Write the plan finishes it at no extra charge.

• Under What to do, each item has a Change line in plain words and an Evidence line linked to the finding, with Ask about this to open the chat on it and Dismiss to set it aside
• Review and apply checks the change with Google, or against the Meta or TikTok ad account, and shows what will change before Apply; pauses and budgets show Applied · Undo
• You do it lists the steps for a change the tool cannot make for you
• Budget figures are worked out from the account's own numbers, every formula says whose figure it uses, and big budget moves and bids always wait for your approval
• Keep watching says what to watch and, under Undo if, when to reverse a change; Changes made keeps one record of what was changed on Google and Meta, with undo
• Your audits lists the brand's audits by day; an audit whose plan was not written offers Write the plan, and an older one with no plan offers Run it again; older Google audits that could not be matched to a brand wait under Not in a brand yet for the owner to place
• Stop ends the run itself, not only what is on screen
• Take it with you exports PDF, Slides, Word, and the plan and findings as CSV, plus a Google Ads Editor file for Google

The chat and your AI teammates read the brand's whole Library

The chat and every AI teammate now read the brand's Library just as the Library page shows it, including what anyone on the brand made. A teammate's conversation and its mail see it too: each turn knows the brand's newest pieces, Email Studio e-mails and uploads, and reads a piece in full when a message links to it or names it.

• A search finds a piece by its words, whatever the accents and quotes
• Email Studio e-mails are found by their name and subject
• A pasted link to a Library piece, an Email Studio e-mail or a tool page is read as that piece, instead of being opened as a web page
• Only that brand's pieces are read, never another brand's

Journey texts reach the person, and SMS wait for their waking hours

Journey SMS and WhatsApp steps now reach the person in every kind of run – audience, event, API and hand-picked – with the phone number taken from the brand's contact record. Marketing SMS keep quiet hours, 08:00–21:00 in the person's own time zone, which the Website SDK picks up from their browser. When it is unknown, the brand's zone is used, then the zone of the journey's delivery window.

• An SMS with a delivery window waits for the first moment both are open, and transactional texts are never held
• The journey's Activity panel says why a text waited, such as Quiet hours for texts (08:00–21:00): their zone unknown – the brand's zone Europe/London
• An e-mail or text step sent to a whole audience fills in each member's own personal fields ({{p.…}}), as a step for one person does; a message that names nobody goes out unchanged

Yearly date: a journey every year on a birthday or an anniversary

Journeys have a new trigger, Yearly date, that starts a journey every year on a date held in the person's profile, such as a birthday or a membership anniversary. Choose the field under The date, then under When they enter pick on the day or a number of days before, at an hour in the brand's time zone.

• Each person comes in once a year on their date, from its first anniversary, and only active contacts enter
• 2026-10-26, 26.10.2026, 26/10/2026 and timestamps are understood, and a year on its own never matches
• 29 February comes on the 28th in other years

Marketing waits while a customer is talking to support, on every channel

While a customer has a conversation open with your support team, the brand's marketing to them now waits on every channel – journey e-mail, SMS, WhatsApp, push, website messages and promo steps – as promo grants already did. An open conversation holds marketing for 14 days from its last message and a pending one for 3, and the newest conversation decides. Service and transactional messages always go.

• The person is recognised by their contact record, e-mail, customer ID or phone
• Held people are counted in the run's summary, such as 2 held – open support conversation, never as failures
• A website message to one person now counts against the brand's weekly marketing budget too

Choose, brand by brand, whether e-mails track opens

Brand settings › Messaging has a new Open tracking switch on the Review & approval card. Track opens with an invisible image is on by default, and can be switched off when your consent rules require it. Clicks are still counted, and every way the brand sends e-mail follows the switch.

Teammates' browser steps finish with a report, or say plainly what stopped them

When the Chrome extension is signed out of the app, a teammate's browser step goes straight to the cloud browser, and the teammate never asks you for a site login because of it. The cloud browser opens public pages without a saved login, pages the goal names keep their exact addresses, and a run stays on the sites it was given.

• If a site blocks the cloud browser – a block page, a CAPTCHA or a sign-in wall – the run stops and says so, with the site's reason and the hint that a member's signed-in browser can open it
• A step is never marked done without a report, and a blocked or failed step can be tried again
• A cloud research report comes back as the step's output

Marketing days, names and STOP replies, read right

Marketing days that move every year now fall on the right date in the chat, Marketing Pulse and campaign seasons – Black Friday on 27 November 2026, Thanksgiving, Mother's Day, the Super Bowl and Chinese New Year among them – and days that belong to one country are left out for other markets.

• An SMS reply with STOP anywhere in it opts the person out; words that can mean something else, such as cancel or end, count only when the reply is short or is about the messages, so "I want to cancel my order" is not read as an opt-out
• When a name is unclear, the gender-neutral forms are used, and a gender imported with the contact, such as F, is respected
• On Czech and Slovak brands, an academic title such as Dr is never printed where the greeting's pan or paní (pán or pani in Slovak) belongs

SEO Meta & Schema results and Library articles read as words

An SEO Meta & Schema result in the chat now reads as words: each option's title and description, whether it fits or Google cuts it, the social card, the structured data, and Open in SEO Meta & Schema. Messages saved earlier read the same way.

• A Writing Studio article in the Library panel and on the Today desk card reads as the article, not as its HTML tags, and a repeated title is dropped
• The chat's done labels read correctly, such as Wrote SEO content
• On SEO Meta & Schema, the snippet and its notes sit side by side only when there is room, so with the docked chat open the notes move under the snippet

AI Visibility: whether AI answers name you, whether AI bots and agents can use your site, and what to do

AI Visibility in Studios is GEO and the AI Tracker as one tool. It asks your questions where buyers ask them – ChatGPT search, Gemini, Claude with web search and Google's AI Mode and AI Overviews – and reads your site the way AI search bots and agents do. A check opens on What to do: one line on where you stand, then one list of what to change, in order, each item with its evidence and an action: a page brief to write from, a review request or pitch for the sites the answers cite, the lines for your robots.txt or llms.txt, or structured data opened in SEO Meta & Schema. It never promises a place in an answer: it shows what was read, and what helps the engines find and read you.

• AI agents: whether an agent acting for a person can find your price, contact or book, and sign up, read without scripts and in a browser, with each control that has no name and any CAPTCHA or login on the way
• Access: robots.txt read for each AI search bot, training bot and agent the way the standard says, and each page fetched as their user agents, so a block or a challenge page shows
• Your files: robots.txt, llms.txt, structured data and IndexNow, each read as it is; a file that is fine says Nothing to change, and lines appear only where something needs changing
• One page, five tabs: What to do, Answers (Mentions, Cited and Share of voice per place, every answer word for word), Site & agents, Setup and History, with every earlier run in one list
• Connected to the chat: Ask about this on any item opens the chat beside the check on that item, and Write it opens the chat with the brief
• Priced on what it reads: about 20 credits for the default check, shown before you start and settled on what it read; Claude is the largest part, and Perplexity is there to tick
• One weekly check for the brand, so you get one alert a week when a bot is newly blocked, a page loses its text, a CAPTCHA appears or you drop out of an answer
• Visits AI assistants sent, from your own site analytics, and the Search Console and Bing AI reports or a server log for what only you can see
• In chat: ask whether AI answers name you or for an llms.txt, and a card shows where you stand

Program calendars: a schedule of challenges or daily offers your team can act on

Marketing Plan Creator now writes a second kind of plan. Choose Program calendar under What it plans, describe what customers do or get each day – challenges, daily deals, an advent calendar, missions – and every day it runs comes back with its own entries: what the customer does, what counts, the window, the reward and where it runs. Download it as a set-up file, turn it into a project, and its live days show on your Calendar.

• Reward for a day's set gives a reward for completing all of a day's entries, and Weekly sets adds a set over each week with its own reward
• Week by week gives an overview, then each week with its days and its sets
• Under Take it with you, CSV, JSON or Copy JSON gives every entry as a set-up file for the platform you run it on; in the Library it is Export CSV · JSON, and the chat offers Download CSV
• Turn it into a project adds set-up, check and pay-out tasks for every week, and the announcements
• Once it is a project, its live days show on the Calendar under Offers & challenges
• Up to 92 days and 300 entries in one calendar, and Runs on picks the weekdays customers can take part
• AI teammates write dated schedules of offers or challenges as Program calendars too, and a calendar a teammate wrote earlier can be rebuilt as one, keeping its days, its sets and its time zone

QR codes in your brand's colour, tracked, and placed on a picture so they scan

QR Code Generator in Studios makes a code for a link, text, an e-mail, a phone call, a text message, Wi-Fi or a contact card, in your brand's colour, with a square or rounded finish and your logo. Every code is saved to the Library as PNG and SVG, and link codes are tracked by default on your brand's link domain, so their scans count in What's Working. Add QR code under a picture in the chat or the Library puts a real code on it and checks that it scans before anything is saved.

• Add QR code finds the picture's code field on its own, or you place the box; drag and resize it to put it somewhere else
• Use a code from the Library, or make a new tracked link on the spot
• Place code then offers Open in Library
• The picture keeps its size, and if a code would not scan the dialog says why
• Any saved code can be put on a picture with Place on an image, and Your codes lists every code of the brand
• Ask the chat for a code, or let an AI teammate's picture step make and place one; placing a code on a teammate's draft updates that draft

Every step finishes whole, and you decide what runs again

Run this step again, in a step's details, now runs only that step and lets the goal carry on from where it stood, and Rewind to here runs that step and every step after it again. A piece that is not whole is never delivered as done: it waits for you, and its record says how many of its items are there. Long tasks are waited for, never started twice, and pause and cancel stop a step that is already running.

• A step does only the work it planned
• Planning that was interrupted starts again on the same goal, from your brief, and the goal page says so
• When a service is busy, the teammate waits it out and says when it tries again, instead of failing the step; picture making has a backup for busy moments
• A batch of pieces shows every one of its items, and later steps build on the whole of what came before
• A teammate's document search finds your uploads by their name and by what is in them

Your words lead a teammate's pictures, and your feedback is kept

What you tell a teammate – the brief behind Make this a goal, a send-back, an answer or a note you kept – now wins over the brand guide in its pictures for what it names, such as a colour, as it already does in the chat. The guide still decides the medium, type, layout and how the logo is used. And feedback you want kept is remembered, for as much of the brand's work as you choose.

• Under the note box, Remember for chooses this goal, everything the teammate makes for the brand, or every teammate on the brand
• Every note reaches every later step whole, newest first
• A send-back on a picture can edit it – keep everything, change only this – at the same size
• What I keep on a teammate's Knowledge tab, and Kept from your feedback in Brand settings › Image Style, list what is kept, each with Remove
• A Keep this card in the conversation offers to keep what you just said
• Kept notes stay through Start over, and when a goal is deleted

AI teammates link to your real pages, and hand over real Page Builder pages

Once the Website SDK is live on your site, every teammate knows your site's real pages: from its sitemap, the pages visitors open, and, when you have them, the latest Site Audit crawl and Search Console. A link to your own site that is not a known page now points to the closest real page, or to your home page. And a teammate's page step makes a real Page Builder page, ready to open and edit.

• The Website SDK page says how many pages teammates know, where they come from and when they were read
• The page shows as the page itself on the Today desk card, on the piece's own page and in the Library panel
• In review, Open in Page Builder and Full screen are one press away

Lifecycle maps on your own stages, and journeys that start on your own events

Customer Journey Mapper now draws a lifecycle map as a document, on the stages your brief names or your business has – three to eight of them. For each stage it gives how customers enter and leave, the events, the messages per channel – service kept apart from marketing – thresholds and suppression, and the metrics, then an event list and the open questions. An AI teammate building journeys can start them on any event your brand registered or received.

• Edit a stage in Customer Journey Mapper, and the map reads as a document in the Library too
• An event your brand does not have is refused, never swapped for another
• Journeys come with waits, conditions, audience checks and exits, and the e-mails earlier steps wrote are attached from the Library
• Every journey a teammate builds is a draft until you approve it

The Library says who made what, and lists each piece once

Work your AI teammates make now carries a byline, such as Graphic Designer · for your brand, on its Library card and in its panel. An image or video you upload is listed once, not twice. And loyalty, referral, win-back, marketing plan and customer journey pieces read as documents in the Library, on the desk and under Read the whole piece.

• A teammate's helper lookups are no longer saved as pieces
• Attempts replaced by Run this step again or Rewind to here leave the Library, and so does a picture version you sent back once its new version lands; nothing is deleted
• A saved Ad Compliance Checker result reads as findings in the Library panel and on its own page, not as raw text

Today puts your decisions on your desk, and Auto-approve lets a teammate's work through

The Today tab on Agents now opens on your desk: the piece waiting for you, large, with Approve and Ask for changes, and with two or more decisions, the rest wait under it in Next on your desk. Below it are Your team, right now, Made by the team, Coming up and The month so far, and Review all takes you through everything waiting. A goal's page is now told in the teammate's own voice, with its phases, Where we are, Ask the SEO Strategist (or whichever teammate it is), Plan & settings and History.

• Auto-approve on a teammate's Work tab is off until you switch it on; then its pictures, copy and pages are approved as they land
• Anything that goes out still follows Auto publish; e-mails to customers, ad money and app changes always ask
• Switching it on says how many waiting pieces it approves right away, and those pieces read Approved automatically
• A piece still waits for you when a brand rule blocks it or its check doubted it
• Approve and start takes an optional note, and a check a goal set up to run on a schedule pauses when the goal does
• A piece whose brand check is still to come says Brand check not run yet

Talk to an AI teammate about new work, and send it pictures

A teammate's conversation now knows your brand profile, what the teammate is responsible for, its goals and why one stopped, its drafts and its ideas. New work for the teammate opens a brief you can edit, and Make this a goal plans it. Replies stream in as they are written, and the conversation opens faster, on its newest messages.

• Paste up to 4 pictures in a message; everyone on the brand can see pictures sent here, and the conversation says so
• Edit and delete your own messages, as in the chat
• A goal that stopped shows Needs an answer, with a box for your answer and Start over
• People appear by name, never by e-mail address, and times are in the brand's time zone

Schedules run on your clock, all year round

Schedules and recurring goals now run at the time in their own time zone, all year, through the clock changes. The chat's task list says each run in its zone, such as Tue, Oct 20, 7:30 AM (Prague time), and the goal list and the Calendar say cadences the same way, such as every Monday at 08:00 Prague time. The chat also knows your local time, so tomorrow at 9 means your tomorrow.

• A time zone the scheduler does not know is refused in plain words: an abbreviation such as EST asks for the place instead, such as America/New_York
• Agents setup fills in your brand's time zone from your browser instead of UTC
• Everyone on the brand sees its schedules in the chat's task list and on the Calendar, with who set each one up

The chat knows your whole brand profile, and remembers for the brand you're in

The chat now reads the whole brand profile – logo, colours, font, style guide and its images, currency, language, goals, channels, greeting, frequency caps and e-mail templates – and says so when a setting is not set. Images, carousels, decks and e-mails made for a brand use that brand's look for everyone with access to it. Enable memory on Profile › Memory now works, and the chat recalls only that brand's memories and the ones for your whole account.

• The chat starts answering sooner
• The side chat uses the main chat's message box, with Ctrl+V screenshots and attachments
• The insights chat on Google Analytics sees the property and the date range on screen

Site tests: A/B, multivariate and site changes, edited on your own site

Experiments now has Site tests. Test your page as it is against up to 7 versions, test up to 3 parts of a page at once, or make a site change for everyone or one audience without a test. Edit on site opens the Chrome extension's editor on your own page, and your edits stay drafts until you publish.

• Your original page is never touched
• Every visitor shown a version counts once, signed in or not, and results come with a 95% interval
• A test has Pause, Stop and Ship winner, and a site change has Turn off
• Edit on site needs the Chrome extension connected
• The old A/B switch in the Personalization panel is retired; site tests take its place

Video narration that sounds like it means it

Narration in Video Studio now gets delivery cues – warm, serious, confident and more – matched to the video's look, its hook and your brand's tone. Captions and cuts follow only what is spoken.

• A promo code written in brackets is read aloud
• A take that would run past its pictures plays slightly faster, at the same pitch
• The words are always your script's own

Share a chat as a page, and find any chat in History

A shared chat now reads like the chat itself, read-only, with its images, Print or save as PDF and Copy link – and only the conversation's words and cards, never what runs behind them. History is grouped by day, with one line per chat saying what came of it, in the Library's words. And the chat reacts to your message with an emoji when one fits, never to a complaint or bad news.

• A password on a shared link needs at least 8 characters, and too many wrong tries pause the link
• Shared pages are kept out of search engines
• Search History, then rename, pin, archive or delete a chat, with 6 seconds to undo
• Deleting the open chat no longer loses what you write next, and opening a chat no longer moves it to Today
• A reply's row is Copy and Export, and a picture keeps its own Regenerate
• The brand you pick is the same brand in every tab, without a refresh

Helpdesk Reporting: any period, an export, and only the people who really answer

Pick any date range for your support numbers with the same calendar you know from Analytics, then take the report with you as a spreadsheet or a PDF. And if someone on your team holds tickets but isn't really on support, such as an owner or a manager, leave them out of the report, so the numbers show the people who actually answer your customers.

• 7, 30 and 90 days stay one click away; any other range is picked on the calendar and kept in the page address, so a shared link opens the same period
• Export gives every table on the page as a spreadsheet that opens in columns in Excel and Sheets, or the whole report as a PDF to share
• The export always matches the screen: the same period, brand and people
• Leave someone out from Who handled it, or tick Counted in helpdesk reports in Team; their replies stop counting as the team's, and they are listed under Left out of reports with Bring back
• Customer numbers never change: every conversation still counts, whoever answered it
• When customers write shows the busiest hours and weekdays, so you can plan who is on and when

Slovak content, brand icons, and hiring an AI teammate from the chat

Slovak is now a content language: choose it in Brand settings and your AI CMO writes native Slovak for that brand. Brands that share one logo can each have their own icon in the brand selector. And an AI teammate can now be hired straight from the chat.

• The app itself stays in English; only what the AI writes follows the brand's language
• Upload a brand's selector icon in Brand settings › Profile
• Who may hire an AI teammate is a tick per person and brand in Team

Site Audit: a crawl of your whole site, and what to fix first

Give it your site and it reads it page by page in the background – up to 5,000 pages – then puts what to fix first at the top: broken links inside the site and to other sites, redirect chains and loops, pages Google can't index and why, missing and duplicate titles and descriptions, pages without an H1, slow and heavy pages, images without alt text, and site-wide checks such as HTTPS, the sitemap and robots.txt. Lighthouse measures the home page and the two pages the site links to most, on mobile. Every number is counted from the crawl or measured by Lighthouse, and the report says what could not be checked.

• Priced on the pages it reads: the price before you start is the most it can cost – up to 5 credits for 500 pages – and what the crawl doesn't use comes back when it ends; a crawl that fails is refunded in full
• A site whose missing pages answer 200 instead of 404 is a finding of its own, because Google can't tell its dead pages from real ones and no crawler can find its broken internal links
• Links to other sites whose servers turn automated checks away are counted apart, as couldn't check – never as broken or as working
• Under the broken links, Find them reads the links from other sites that still point at your dead pages – 2 credits
• Broken Link Checker and Page Speed have their own tiles in Studios; Page Speed measures any page with Lighthouse for 1 credit an address
• Ask chat to audit your site or find broken links: a card under the reply follows the crawl and fills in when it finishes, and the bell rings. Asking what it found costs 0 credits
• The SEO Strategist reads the latest audit, and a technical fix it proposes names the findings it fixes
• Your audits keeps every audit of the brand, opening one again costs 0 credits, and each downloads as CSV, PDF, Word or Slides

Helpdesk knows who a site chat customer is once they log in

A visitor who logs in on your site while they are chatting is attached to the conversation within seconds, without writing again, and the whole conversation stays. Helpdesk shows it as a line in the thread: Customer logged in as … – identity verified. A site that names its logged-in customers without a signature no longer leaves them anonymous: Helpdesk shows them as not verified, so your team knows to confirm who they are, and a new Identity tab in the Site chat studio says how your site's customers arrive and how to sign them.

• A customer named without a signature is never attached to a customer record, sent a reply e-mail, or shown History or their chat on other devices
• With Require signed identities on, unsigned names are ignored and the chat stays anonymous, as before
• A login attaches a chat the visitor wrote in within the last hour, on the chat's next check; an older chat becomes theirs with their next message, so the next person at a shared computer never inherits one
• The Identity tab shows whether customers arrive signed, unsigned or with a signature that doesn't match, and why, with the secret and the switch beside it; changes save at once, not with Publish
• While the chat is on and customers arrive unsigned, a line over the studio's preview says so, and the brand's administrators see one line across the top of Helpdesk's Chat view, and over the brand's open chats, with Set up signing
• Identity verification in Website SDK now says what signing does for Site chat, and its code signs every identify() call

Helpdesk: tickets find their owner, files travel both ways, and nothing slips by

Open a ticket nobody holds and it is yours, with a line in the thread so your team knows. A new ticket now plays a clear sound on every channel, photos and files your customers e-mail land in the ticket, and you can attach files to your replies. Messages from your website's contact form now go to the person who wrote, not to your own address.

• The first person to open a ticket gets it; a ticket a teammate already answered or holds is never taken from them, and Nobody hands it back
• A clear sound for a new ticket and a softer one when a customer writes back, with volume and desktop alerts under Sound on the queue
• Attach up to five files to a reply – PDFs, pictures, Word, Excel and PowerPoint – up to 2 MB in all
• Photos and files customers send to your connected mailbox show in the ticket, ready to open
• E-mails read the way the customer wrote them: mailbox warnings and link clutter are tidied away, and e-mails written only in HTML show their words
• A contact form message that arrives from your own address with the customer in Reply-To becomes that customer's ticket, and your reply goes to them

Czech and Slovak SMS fit in one message again

One accented letter turns an SMS into the Unicode format, which fits 70 characters instead of 160 and roughly doubles what you pay. Czech and Slovak texts now go out without accents by default, so a normal message fits in one SMS, and the counter shows the text exactly as it will be sent, with the real name filled in.

• Keep accents on any single text with one click under the numbers in SMS Studio; the counter shows what each choice costs
• The counter counts the message your customer receives, not the placeholders you typed
• Journeys follow the same rule automatically
• Links and phone numbers are never touched

Your brand's name works in every message, and the preview line reaches the inbox

Write {{brand.name}} once and it becomes your brand's name in e-mails, SMS, push, website messages and journeys, in the subject and preview as well as the body. The preview line your e-mails carry now shows in the inbox right after the subject, and the AI writers know they can use your brand's name this way.

• Previews of your e-mails in the chat and in Journeys show your real brand name and logo
• Whatever is stored about a customer always goes out exactly as written, so nothing in their details can change your message
• Greeting examples in Brand settings follow your brand's language: Czech, Slovak, English and more

Watch your video come together, step by step

Ask the chat for a video and the Video Studio card appears straight away, saying what it is doing right now: writing the brief, writing the scripts, planning the scenes, then rendering. One timer, an honest estimate based on real videos, and you can leave at any time – the finished video lands in your Library.

• Watch it render appears only once the render has really started
• Come back after a reload and the card picks up at the right step
• If something stops, the card says what happened in one sentence and how to carry on
• The same card in the side chat

New people start without marketing consent, wherever they come from

A person added to your contacts now starts as Not granted for marketing e-mail unless consent was really given, on every road: a landing page, a survey, a Shopify or Stripe sync, a file you import, an audience built from your data, the journey trigger API, a customer typed in by hand or a first e-mail to your support. Service e-mail still reaches them, and their record shows it under Compliance. Someone already in your contacts keeps what they had, and consent they gave you before is never undone.

• A landing page's consent box now counts: a tick is recorded with the page and the words shown, and a new lead who leaves it unticked is not e-mailed for marketing
• Shopify buyers subscribed to your e-mail marketing in Shopify arrive with that consent
• A consent column in your file, such as email_consent or Shopify's Accepts Email Marketing, sets each new person's consent; the import says whether it found one
• For someone already in your contacts, a yes from your data (a file, Shopify, the API) counts only when Not granted is all that stops them, never over an unsubscribe
• The journey trigger API takes consent: { email: true } when your own system holds the person's consent
• An e-mail added to a customer who has none starts Not granted unless you tick Marketing e-mail in the same save, and correcting the address of a Not granted customer keeps them Not granted

The chat's first page now tells you the next move

A new chat opens on at most two cards drawn from your brand's own numbers and work – revenue that moved, customers slipping away, your plan's next step, or a first job for a new brand – each with one button that prepares the request for you. Nothing is sent or charged until you press send.

• Under the cards, one line of what was made this week and any draft waiting for your yes
• After a day away, the chat opens here with a link back to where you left off
• Sending a message now glides smoothly into place, and a long reply never pulls you away from what you are reading
• Paste a screenshot straight into the chat with Ctrl+V or Cmd+V

Coach me now lives in the chat

Say the outcome you want and by when, and your AI CMO writes a plan and takes you through it one step at a time, in the same conversation. Start one with Coach me under the message box, or from the card the chat offers when you ask for a plan or for steps. The plan opens beside the conversation, a finished step is one press, and the plan's chat is in History whenever you come back, and on Home while the plan is in progress. The separate Campaigns page is retired: its old address opens the chat, and the goals your agents run are on Agents, as before.

• You see what planning costs before anything is charged, and it is charged only once the plan is written
• Planning is its own line in your credit history, Coached plan, and never counts toward your AI team's daily or weekly limits
• Channels are optional: leave them to the plan, or choose from the ones suggested for your outcome
• Everyone on a brand shares its plan and its progress, each in a chat of their own
• End or delete a plan from its menu. The chat stays
• In History a plan's chat says where the plan stands, and a Plans tab lists them
• A plan's steps are on the Calendar as your own work, beside your tasks, and never overdue
• On the Calendar each step sits under the channel of its work, and Plans in the filter row shows them on their own
• A plan nobody has touched for a week gets one note in the bell; a week later its next steps leave the Calendar until you pick it up

Connect a platform without leaving the chat

Ask your AI CMO to publish a post, read your traffic or read your ad results, and if that platform isn't connected yet, it now stops the step before anything runs, says so in one small card and offers Connect right there. It is the same sign-in as the Integrations page. Once you're signed in, a post goes straight to its publish question, so publishing takes two presses: Connect, then Publish. A read waits for one press on Carry on, with its price on the card.

• A held step isn't charged, and runs only once the platform is connected and you say yes
• A login that stopped working shows Reconnect
• With several brands you choose which brand the account is for, with the chat's brand already selected
• Email, SMS, WhatsApp and app store keys open Integrations in a new tab at the right form
• Not now closes the card

Actor Library: every cast member is a character sheet – one person, drawn from every side

Describe a character in a line or two, or upload a photo of someone you already work with. A casting director writes them in full – their face, a signature look people remember, a head-to-toe outfit in your brand's colours, how they come across on camera and three expressions – and you can change every word before anything is drawn. Then one picture draws them from every side: five full-length turns above three portraits. Every video you cast them in receives the eight views and a portrait drawn from them, so the face, the build and the outfit carry from shot to shot.

• From a photo: their face stays theirs, their outfit follows the photo unless you say otherwise, and your confirmation that you may use their likeness is recorded with its date
• Each new character is made unlike anyone already in your cast
• About 1 credit to write a character and 5 for the sheet and portrait; if the picture does not come back, the credits do
• Change who they are on their page at any time, and Redraw sheet draws the same person again
• Actors made before today keep working, and Draw their character sheet gives them one

Teach your AI teammates from your own material

Every teammate except the support employee can now be taught, on the Knowledge tab of its profile: how it should sound, how the work is laid out, what to always or never do, and what it should know. Give it a file, pieces from your Library, a web page or a few lines of your own; it writes down what it took from them, and nothing counts until you press Use this. From its next piece of work it writes with it – in the plan, in each step, in the tool that writes the piece and in scheduled runs.

• A report layout taught as a spreadsheet comes back as a table with exactly those columns – every table in the Library now has Export CSV
• The Social Media Manager can keep a lesson to one platform, and its lessons shape the replies the Helpdesk drafts on social threads
• Each lesson shows how many pieces of work it has shaped, and a result made with one says so on its Made by line
• Studying is priced on what it reads: the form shows the most it can cost, the rest comes back once it has read, all of it if it cannot be studied
• The answers you gave when you hired a teammate now reach the writing too, not only its plans

App Store Optimization: what to change in your App Store and Google Play listing, read from the stores themselves

Add your app's App Store link or Google Play package, the markets and the searches it should be found for. The stores are read first, free – your listing today measured against each store's limits, where the app shows for each search and the apps around it. Then, for 15 credits, you get recommendations first – in order, each tied to what was read – and three new listings for each store and market, every field measured the way the store counts it and checked against your brand's rules.

• Where the app shows for each search, as a grid beside the apps around it, market by market
• Search demand only with its source named – Apple Ads or your Play Console reports when connected, an estimate for the United States in English, Google web searches elsewhere, never passed off as store searches
• What recent reviews say, in themes quoted from the reviews; today's screenshots beside what shoppers see first; a screenshot plan and store tests to run
• A PDF, slides or Word report with your brand's logo and the app's icons
• Nothing reaches a store unless a person sends it: with App Store Connect or Google Play connected, the option you pick goes into the App Store version you can still edit, never submitted for review, or to Google Play, which reviews it before it shows
• With the stores connected, their reviews reach the Helpdesk every six hours: your support employee drafts for reviews of 3 stars or fewer, and a person sends every reply – at most 350 characters on Google Play
• Mark it live, then read the positions again from Your listings, with the stores' own numbers either side of that day when connected
• The Google Play feature graphic (1024 × 500) in the Image Generator; chat and your AI team run the same tool

Image Generator: on-brand images at each placement's exact size, on Nano Banana 2 or Muse

Describe what the image shows, pick where it goes – an Instagram post, a story, a shared link, a website hero, a display ad or your own width and height – and it arrives at exactly that size, in your brand's image style. Choose the model: Nano Banana 2 for images that carry words, or Muse from Meta, which renders at your exact size and is strong on photography, people and scenes.

• Words only when you give them – a headline, a supporting line and a button, set exactly as written
• Your brand's look, colours and mood shown on the page, with the logo when it fits, on or off
• Up to six references – a product, a person or a place – from an upload or your Library
• Up to four images at once, each shown as it is ready, with the idea it was built on
• Every image saved to your Library; 3 credits each, and one that fails costs nothing

SEO Meta & Schema: titles that answer what people actually search, measured the way Google cuts them

Pick a page and it is read the way Google fetches it – its title, description, tags and text – and, with Search Console connected, the searches it is really shown for. You see today's result as Google sets it, what stands in its way, and three new title and description pairs, each from a different angle and written only from what the page says. Once the new one is live, Search Console shows the clicks before and after.

• The pages Google shows often and people rarely click, listed from your Search Console
• Widths in pixels, not characters – Google cuts by width, so a wide word costs more room
• Each option shows which of the page's real searches it answers, word for word
• It flags what holds a page back: a noindex, a canonical pointing elsewhere, a past year, the top search missing from the title
• Tags ready to paste, the card a shared link shows, and structured data that never adds a rating or price the page does not show
• No score: the result is measured after the change, 28 days either side
• Chat and your AI team run the same tool from just a page address

Subject lines in Email Studio: checked as you type, tested for real

The subject and preheader are checked while you write them – what a phone cuts, words spam filters weigh, a fake "RE:", a placeholder left in, a preheader that repeats the subject – and the e-mail shows as a phone inbox lists it. Your own subjects that got opened most are listed beside them. To find out which of two subjects works better, Send now tests a second subject: half the audience gets each, and Experiments shows the opens and clicks.

• No predicted open rate – nobody can know one before the send
• Suggest writes alternatives; test any of them against the subject you have
• Any journey e-mail step can carry its own subject, shown on its card
• The standalone Subject Line Tester now lives here, where the subject is written

AI Tracker counts how often AI assistants name you

The questions a buyer asks about your category go to ChatGPT, Gemini, Claude and others exactly as a buyer would ask them – with nothing about you in the prompt. Every answer is then counted: whether you are named, where in the list, your share of voice against the competitors you name, and the sources cited. What the assistants get wrong about you is checked against your profile.

• Runs are kept per brand, with the trend from one run to the next
• Every answer is there to read, not only the totals
• One credit per answer, shown before you start

Ad Compliance Checker: the platform's rules as you type, then a review

Write or paste an ad for Google, Meta, LinkedIn or Amazon and the rules that can be counted are checked as you type – lengths, restricted areas, personal attributes on Meta, claims, trademarks and your industry's words. A review then reads what rules cannot, and gives rewrites you can apply. The verdict says what would likely be disapproved and why, without a score.

• The ad shown as the platform shows it
• The same review in chat and for your AI team
• Reopen any check from your library

ROI Calculator works from your own spend and revenue

Return on ad spend, profit after your margin, the break-even return, what a new customer costs and how many orders pay it back – per channel, from the same numbers Customer Intelligence shows. Without a margin it shows revenue returns only and says so, rather than assuming one.

• Channels that earn without recorded spend – e-mail, organic – are shown apart
• Move budget between channels and see the change, with the assumption stated
• Your own numbers work too, when the data is not connected

PPC Campaign Builder: Google search campaigns on what one market searches for

Choose the country and the language people search in, and the campaigns are built on real searches there – keywords, match types and ad groups decided from the data, the budget in your brand's currency, and ads checked against Google's rules. With your Google Ads account connected, the searches that already converted come in and the ones that only cost money stay out. Start from a saved strategy and its positioning, offer and audiences steer the ads.

• Sitelinks, callouts and snippets point at real pages and facts from your own site
• What was left out, and why
• A file for Google Ads Editor with every campaign paused, so nothing spends until you switch it on

Flash Report plans are your own: your lines, your months, your industry

Upload the budget you already have and every line comes through in your own words and sections – GGR, NGR, bonuses, licence fees, active players – over your own months. A launch plan numbered 1 to 12 goes into the calendar from the month you name, across the year end, and approving it freezes exactly those months.

• Stacked tables on one sheet are read table by table; hidden working sheets are left alone
• Each line is compared with the report's own figure when it is the same thing, or with the figure your systems send for it – never worked out from other data
• Send those figures daily or monthly through the data API, beside your events and plans
• Plans and the report speak your industry: a casino reads deposits and first-time depositors, a lottery ticket sales, never orders
• Each line is recognised as your industry's usual line – registrations, first-time depositors, bonuses – in your own words
• Monthly plans are read by day, so a month to date, a week or any custom period has its own plan
• The report shows as soon as a plan exists – before any actual, and for a month that has not started
• Your plan, line by line, in the report and at the top of the CFO PDF and PowerPoint

Camera path: place the cameras, and the film flies through them

A new look in Video Studio where you direct the camera. Your product stands in the middle of a simple stage with a ring of cameras around it; click them in the order the film should pass them, and each one takes a number and the path is drawn between them. Describe the scene – where it is and what happens – and Render turns the move into one continuous shot of your product there, following the path you set.

• The stage is only for placing the cameras: the place in your video is the one you describe
• A bottle, can, jar, box or tube stands in for your product, or a person, so every camera frames it at the right size
• Each camera is Wide, Medium or Close, and a small frame shows what it sees in your video's shape
• Play the path before you render, to watch the whole move
• Made on SeeDance 2.5, or MiniMax H3 when a cast member's exact face matters, with music made for the look

Put yourself in your videos, and SeeDance 2.5 joins Video Studio

Beside Actor, a new Person tile adds a real person to a video from one photo of their face – you, your founder, anyone who agreed to it. The studio reads the photo and writes how they look, you confirm you have the right to use their likeness, and they're in the film. They join your brand's cast, so next time they're one click away.

• Reading the photo costs about 1 credit and adding the person 2; a photo with more than one face or with a child in it is declined, free
• SeeDance 2.5 is now in the model list: ByteDance's newest, holding one take for up to thirty seconds, at 480p or 720p, and it speaks your script

Edit an area: fix one part of a picture and keep the rest exactly as it is

When a picture is almost right – a word spelled twice, a line of text running over a photo, an icon you'd rather swap – paint over that part and say what should change there. Only the painted part is redrawn. Everything else stays the original picture, pixel for pixel, at its full size, so fixing one thing leaves the rest untouched.

• In chat, choose Edit an area under any picture or in its full-size view
• In Image Center, select a picture and choose Edit an area in the tools
• Three brush sizes, with Undo and Clear if you overshoot
• Regenerate tries the same painted part again, and a failed edit is refunded
• Every picture in chat now shows its Download button, however short the message that comes with it

CFO pack: a PDF and a PowerPoint you can put in front of the board

The CFO PDF and CFO PowerPoint from a closed Flash Report have been rebuilt from the ground up. Both open with the month's headline figures and your management commentary, then take you through the scorecard against budget, forecast and target, the trend, what your ad spend returned, where revenue came from and how your customer base moved. Every chart has a labelled scale, every table reads cleanly, and the PDF and the deck always show exactly the same numbers.

• Drawn in your brand's colours from Brand settings, adjusted where needed so every chart reads clearly on the page
• Figures are written the way your currency is – a Czech pack reads 249 073 and 1,89 % next to Kč
• Rates such as click-through move in percentage points, and a change under 1% reads as flat
• New, retained and lapsed customers now compare with the month before, like every other figure
• The PowerPoint is fully editable, with native charts and tables and speaker notes on each slide

Flash Report: see each month against your budget, your forecast and last month

The Flash Report now measures your month against the numbers your finance team works to: the budget you agreed for the year, the forecast you keep up to date, and your target. Approve the budget once and it stays fixed, so every month is judged against the same plan. There is nothing to retype – upload the budget spreadsheet you already have, in its own layout, and check how each row was read before anything is saved. A month that is still running is compared with the same days of last month, so it no longer looks like a decline until the month is over.

• Green means better and red means worse – when spend comes in under plan, that is green
• Close period asks which finished month to close and shows what will be frozen before it does
• Your data team can send the budget and forecast straight from your own systems, through the same connection as your data
• The report loads several times faster

The right brand, every time

If you run more than one brand, two things now follow the brand you are working in. Scheduled social posts publish from the account connected to the brand you scheduled them for, and Google Analytics Insights shows the right brand's data when you switch. And the owner of a brand that has closed Flash Reports or marketing mix models can now delete it – a record of each closed report is kept for your audit trail.

Google Search Console, built into your SEO work

Connect Google Search Console and your SEO work starts from where you actually stand in Google. Keyword Research shows your current position and clicks next to each keyword. The SEO Meta Generator starts from the pages Google shows often but people rarely click. An SEO goal shows how its pages did in the four weeks before the goal began and in the latest four weeks, and Link Building shows the clicks each article earns. Chat and your AI teammates can read it too, so you can simply ask how a page is doing.

• The SEO Strategist checks your real rankings, your competitors and today's search results before it suggests anything, and each idea shows the figures it is based on
• Before-and-after results show what changed on the page – they never claim what caused it

Brand checks run once, on finished writing

Your brand rules are checked when a piece is finished, not while it is still being written, and the same text under the same rules is checked once – opening it again, or a teammate looking at it, shows the same answer without another check. Articles started from a project task are now only the article: the task's other steps, like publishing or linking to it, no longer end up as its last paragraphs, and the document takes the article's own headline.

• A draft deleted in Writing Studio or the Library leaves its task's work and shows in the task's activity – as a deleted page already did
• A task's activity names each draft as it is now called, not as it was first saved

Article pictures at the size the article needs

Pictures in articles – in Writing Studio, in chat and in the articles your AI teammates write – are now made at exactly the size the article shows them, with no words painted in. The original stays in your Library, and the article loads small web copies of it, so it opens fast wherever you publish it.

• In Writing Studio, choose a size per picture: what fits the article, a common size or your own
• Make pictures web-sized gives older articles web copies – nothing is generated or billed

Pages that look like your site – header, footer and all

The Page Builder now reads your website the way a designer would: it photographs your homepage and follows what is actually on screen – your colours, your type, your buttons, how your sections sit, the kind of pictures you use. It reads your whole homepage and the pages your menu leads to, like Chat does, and builds your own header and footer once, to put around every page – or just your logo, or nothing. Pictures are made at exactly the size their place needs, with no words painted in, and the page loads small web copies of them instead of the full-size file.

• Page → Your site shows what was read, with Read it again
• Choose a size per picture: what fits its place, a common size or your own
• Make pictures web-sized fixes pages made before – nothing is generated or billed

Page Builder – a studio like the rest: review, quality, results

The Page Builder is rebuilt as a studio in the same shape as Email and SMS Studio. Drafts save themselves, and a live page changes only when you update it. Each section can be edited directly – its words, where every button goes, and its images: generate again, pick from your library or upload. Quality lists what is still open before a page goes live, Review carries the approval and comments your e-mails already use, and Results shows views, visitors, leads and where they came from. Pages are written in your brand's language, a page from chat reads your site just like one built here, and the section layouts no longer carry sample quotes or prices.

• A page stays with the brand it was made for; its form joins only that brand's journeys
• An optional consent box on forms and an optional notice before pixels load – both off until you turn them on, in your own words
• Previews inside the app no longer count as visits
• Build the page on a task opens the Page Builder with the task's brief filled in; the page comes back to the task as its draft

Page Builder – any page, and a way to hand it to your developer

The Landing Page Builder is now the Page Builder. Describe the page and it builds for its job: a landing page that converts, a proof or resource hub buyers can inspect, a comparison, an explainer, an event page – on your brand, in your site's own styling. When someone else runs your website, Send to developer e-mails them a private link with a preview, the page as one HTML file with its forms and tracking built in, and how to put it live, including the exact DNS record when you have connected a domain. If your own team builds its pages, a page task sends them the brief instead – what the page is for, what goes on it, the project behind it and the date – without generating anything. You see who got it and whether they opened it, on the task too.

Banner Studio: every size, in your brand, ready for the ad server

Banner Studio is rebuilt. Start from a canvas at any size, a template, or a brief in your own words – choose where the words sit and where the button goes, and the banner is built as editable layers on a picture made at that exact size, from a 728 × 90 leaderboard to a 1080 × 1920 story. Your brand colours, typeface and logo lead everywhere; the logo is yours to switch on and place. One design becomes every size you need, each laid out for its shape rather than stretched. Export PNG, JPG or WebP at the exact pixels, or an HTML5 package per size with the animations, the ad size and a clickTag built in – made for Google Ads, Google Ad Manager and DV360.

• Chat and your AI teammates make banners in Banner Studio too: the sizes you name, the words as editable layers, a preview in the conversation and a link to open it
• Banner Generator and Image Banner Generator are now part of Banner Studio – their old links open it
• Everything the editor did stays: layers, the animation timeline, PSD import, your own fonts

Images at any exact size – in Image Center and in chat

Image Center's Create makes an image at the exact width and height you set, from 64 to 4,096 pixels, following up to four reference images. Smart Resize now rebuilds one image for every platform's formats and your own sizes at their exact pixels, instead of cropping, with a free crop still there when that is all you need. Ask chat for an image at a size – "952 × 458, like this one" – and it arrives at exactly that size. An image carries no words unless you ask for them.

• Every tool charges exactly the price on its button – Combine no longer charged twice, and nothing is charged from the browser
• Everything you make lands in the brand's Library
• Resize keeps proportions locked unless you choose to stretch

A strategy now makes a call – and shows its working

The Strategy Creator writes a strategy the way a strategist argues one. It opens with a diagnosis – where your brand stands and the one crux worth solving, with the evidence – then the bet that attacks it: its signature move, why now, and what the strategy deliberately won't do. Every tactic says how it serves that bet and carries a button that starts it with the right tool. It reads more than your brief: your customer data when it is connected, the searches your website already ranks for in your market and when your category peaks, what the competitors in your profile are found for, and your last strategy, so a new one builds on it. A brand on its first day still gets a committed strategy – the numbers it relies on are listed as assumptions and week one gets your measurement live.

• Every section the strategy writes is now on the page: who it is for, the offer and funnel, creative directions, measurement, experiments with kill criteria, assumptions and open questions
• Answer the open questions and Sharpen the strategy writes a new version built on this one; your answers are kept for every later version
• Chat opens beside the strategy on the Strategy tab, already reading it – and stays closed while you set up

Promotions budgets, your own channels, and twelve ways to focus

The budget you give a strategy is now the total, and Of which promotions sets the part that goes to your customers as value – bonuses, discounts, coupons, free shipping. The strategy splits media from promotions and plans each promotion: what the customer gets, who qualifies, when it fires, its cap and the result it must produce, built on the promos you already run. Organic strategies can carry a promotions budget too. Add channel adds one the list does not have – podcast ads, retail media, your own – and the strategy budgets for it like any other.

• Focus now has twelve options in four groups, each explained: the whole funnel; paid media, search, AI search, content and thought leadership, social and creators, partnerships and affiliates, account-based marketing; lifecycle and CRM, retention and loyalty; brand, launch and market entry

Strategy, plan, loyalty, referral and win-back share one setup – and plans start when you say

All five tools now set up the same way: your choices in a panel on the left, the result taking shape on the right, read back as the document's first line with a list of what it will contain. Options that need explaining – program type, who gets rewarded, the win-back approach – say what each one means. Win-back leads with your real lapsed customers and the revenue at risk. The Marketing Plan Creator lets you pick the plan's start date instead of always starting tomorrow, and a plan built from a strategy now carries its own channels, its promotions and its signature move, scheduled from the first weeks.

• Fixed: a plan made from a monthly budget read it as the budget for the whole plan
• Fixed: the Make it run buttons under a loyalty or referral program did not open the chat
• Budgets in loyalty, referral and win-back exports are in your brand's currency

Ads from the chat lead with what is worth promoting

Ask the chat for an ad and it now decides what to promote before it decides how the ad looks – the one reason your audience would click, not the detail that is easiest to draw. Point it at one of your pages and the ad leads with what that page sells, its main promise, so the click lands on the same message; the page's features, figures and quotes become the proof for it. Ask for another one in the same conversation and it takes a new angle instead of repeating the last ad's headline, while keeping any look you asked it to keep.

A generated image now costs 3 credits

An image from the image generator or the Image Banner studio used to cost 15 credits. It now costs 3 – the Manager plan's 5,000 monthly credits cover about 1,650 images, the CMO plan's 10,000 about 3,300. Asked for in the chat, the message that asks is priced on its own, like every chat message.

Chat, AI writing and your AI teammates are priced on what they use

Every chat message now costs credits for what it actually uses. The first message after an hour away loads your brand's full context and comes to around 12 credits; every message after it within the hour reuses that context – a reply is usually 1 to 3 credits, a message that makes something with a tool a few more, plus the tool's own price. The AI writing and analysis that had no price of its own – landing page writing, Email Studio, reading a website into a brand, briefs and insights – is priced the same way. Planning a goal for an AI teammate comes to around 30 credits, charged once the plan is written, and the teammate's own thinking as it works is priced like chat. Video has new rates too: a second costs 1 to 8 credits depending on the model and resolution – SeeDance at 1080p is 4.5, Wan at 720p is 2. A cloud browser task reserves up to 40 credits and settles on what it used. A message or task that fails costs nothing, and every charge shows in Usage.

Watch the cloud browser work, live in the chat

When a task runs in the cloud browser, the chat now opens the live browser window the moment it is up and keeps it in view while the task runs. What the agent is doing right now is written over the window, its steps stream underneath as it goes, and if a page asks you to sign in, you do it right there. The picture is taken by us straight from the browser every few seconds, so it shows the page even on sites where the vendor's own view stays blank. When the task is done, the reply keeps the page the browser ended on. The reply also names what the run actually cost. If you chose the cloud browser in the composer, the chat no longer asks you to confirm it each time.

Choose where browser tasks run – your browser or the cloud browser

Until now the cloud browser only stepped in when the Chrome extension was not installed. You can now choose. In chat, the browser control in the composer has three settings: Automatic, which uses the extension whenever it is connected; Your browser, which always works through the extension with your own logins; and Cloud browser, which always runs in a browser on our servers, leaving from your brand's country, and keeps going when your laptop is closed. On Agents, every specialist has the same setting under Settings & details, so a teammate that reads competitor sites can live in the cloud while a teammate that posts for you works through your browser. X, Instagram, LinkedIn, Facebook, TikTok and YouTube always run through your browser, whichever you choose: they challenge a cloud browser at login.

The cloud browser is priced on what it uses

A cloud browser run used to cost a flat 40 credits however short it was. Now it reserves up to 40 credits when it starts and settles on what it actually used when it ends – most tasks come to a few credits, and the rest of the reserve goes back to your balance the moment the run finishes. A run that fails, or that you stop, costs nothing. The same meter prices browser work in the Chrome extension, so a task costs the same wherever it runs. Every run shows its own line in Usage: what was reserved, what it used, what came back.

Agents shows you what it made and what it needs – and you can talk to a teammate in the app

The Today tab on Agents is now two lists. Needs you holds everything the team is waiting on a person for, with the button that resolves it on the card: a question a step stopped on, a plan the reviewers held, a login it needs, a piece of work waiting for your approval, ideas waiting for your pick. Done for you holds everything the team has made, as the same cards you know from the Library – the title, a miniature, which teammate made it and when – and opens over the page in the same reader as the Library. And every specialist can now be talked to where it works: ask about its work, where a goal stands, why it chose a step; write status, pause or resume.

• The headline reads from the same list as the cards, so it never says the team is choosing what to do while a question waits – including on a paused goal
• A question a step stopped on is written in plain words – the step by name, what went wrong, and retry, skip or your own answer – with Add credits, Raise the daily limit or Add the login right there when that is what stopped it
• Every result says which teammate made it: the specialist, or the goal it ran under; nothing waiting on you is ever cut off the list, and earlier results load a page at a time
• A goal page opens with What it will deliver, the mix of work the plan produces and what is done, and What it needs from you, every login and connection with the button to add it; each finished step shows the result by its own title
• The conversation is a thread: the teammate speaks first, three questions to start from, and a message about a goal that is waiting on your answer is taken as the answer, and it says so; an answer it writes costs 1 credit, status, pause and resume are not charged, and nothing is approved or published from a conversation
• Everyone on the brand sees the conversation, each message carries who wrote it, and the e-mail address under the name still works the same way

A support employee for your helpdesk – it learns your desk first, then answers only what you allow

Hire the Autonomous Customer Support Employee on the Agents page and it starts by learning. For every message that arrives it writes the reply it would have sent, shows it to no one, and compares it with what your team actually sent. When enough of them match it says so, in plain numbers, and you decide whether to make it a Copilot – its reply then waits in each conversation for someone on the desk to approve, edit or discard. On its own, it answers unattended only the kinds of question an administrator has switched on, one topic at a time. It never moves itself forward, and it never switches a topic on.

• It answers from your saved replies and the documents and pages you add to Knowledge, and every draft shows which of them it used; a colleague's past answer can shape a proposal, never an unattended reply
• What it does not know it writes down: every question it could not answer is listed for your team, and one answer teaches it from then on
• Before anything goes out unattended a second check reads the reply against its sources – one it cannot support waits for a person, and so does every conversation after three answers in a row
• Self-exclusion, signs of harm, chargebacks and threats of legal action never reach it; they go straight to a person
• Names, e-mail addresses, phone numbers and account, card and document numbers are replaced with placeholders before a model reads a message, and nothing runs until an administrator has confirmed the addendum – what, when and who stays on record
• Its work costs credits: 1 for a reply written while learning, 2 for a proposal, 3 for a reply it sends itself. Pause all pauses it too, and the desk is told when credits or a limit have stopped it

It can look things up in your own systems – and be shown how a task is done

Some answers live in a back office no integration reaches. With Lookups on, the support employee opens one customer's page in a system your team already uses, through the browser extension of someone on your desk, reads what you told it to read and answers with it. It runs only the lookups your team wrote down in plain words, and it is told to read and never to save, approve, refund or change anything. Because a real identifier is involved, Lookups have an addendum and a switch of their own, off by default. Training is the first step of teaching it procedures: record your screen doing the task, it writes down the steps it understood, and you check and approve them, version by version.

• While learning it names the lookup and the desk gets a Look it up button; as a Copilot it looks first and then proposes; on its own it answers with the fact only when the customer's identity is verified
• Reading only is an instruction, not a lock – give the account it uses read-only rights in that system
• A lookup costs 5 credits, given back in full when it could not be completed
• Training costs 5 credits per started minute of recording (at least 10) or 1 per screenshot (at least 5); record it with a test customer, and the recording is deleted once you approve or discard the steps
• Approved procedures are not used in its replies yet – that is the next step
• Where a part of it is not switched on yet, the page lists what is left to do, in order, with the button for the next step right there

The Office – limits that fit your plan, rules that deliver, and a Save you cannot miss

Three things on the Office tab of Agents. The daily and weekly credit limits of a new brand now start from its plan – the monthly allowance divided by thirty, rounded down – instead of the same 50 a day for everyone; limits you have already saved are not changed. Automatic rules now run their tool the way your AI team runs a step of a goal, so the Weekly Social Planner really does leave five posts in your review inbox on Monday. And the moment you change a setting, a bar at the bottom of the screen says so, with Save and Discard.

• A rule that could not run is tried again after 30 minutes, three attempts in all, and then tells you plainly – it no longer shows up as a draft waiting for review
• Nothing is charged for a rule run that produced nothing
• The support employee keeps answering customers when the Office limits are reached unless you choose to hold it to them; either way the desk is warned at 80% and when it stops

Chat shows what it made – as the thing itself

What a tool produces in chat now arrives as what it is: an e-mail as an e-mail, a post as a post, an ad framed by the network it was written for. Ask for three e-mails and you get three cards, each its own, with a progress card that follows the turn while they are being written. A shared chat link and the Library draw the same cards the chat does.

• A step that is still running no longer draws an empty card
• A run your browser lost track of is settled from the server's own record instead of hanging
• A tool that goes quiet halfway through an answer no longer blocks the conversation

A personal e-mail is plain text – no template, no invented footer

Ask chat for a cold note, a sales follow-up, an introduction or a partnership ask and it now writes a personal e-mail: plain text, short, in your voice, with anything it does not know left in [square brackets] for you to fill in. Until now every e-mail came out as a branded mailer, footer and all. Newsletters, promotions and everything you send to subscribers stay designed and sendable as before; a personal note is copy for your own mailbox, saved to the Library as an outreach message.

The Library, rebuilt – everything made for a brand, in one place

The Library now lives at its own address and reads as one thing: a headline with the real count, a strip by kind – content, assets, research, plans, files – search that looks inside the text, and cards that show the piece itself rather than a tool's label or a block of JSON. It belongs to the brand: a teammate sees the same Library, every item signed by its maker. Open a piece where it lives, hand it to chat, star it, move it, rate it, comment on it.

• Uploading files works again, and a document uploaded here is readable in chat the moment it lands
• Search finds a strategy by a word inside it, not only by its title
• Old links and bookmarks still open – they land on the new page with the same filters
• Nothing in the Library is called a campaign any more; a campaign is what you run on the Campaigns page

Chat reads the documents you attach

Attach a brief, a deck, a spreadsheet or a report with the paperclip and chat reads it before it answers – PDF, Word, PowerPoint, Excel, CSV, text and Markdown. Until now a file arrived as a name only. The chip on the attachment says what was read, a scanned PDF is called out as not readable rather than guessed at, and every document you attach is saved to the Library for the brand, where teammates can open it and later conversations can ask for it by name.

• Long documents are read in part – chat says when one was cut and asks which section matters
• The paperclip's image and content pickers read the brand's Library properly – searched on the server, real titles, no tool labels, no raw JSON, identical scrapes folded into one
• Open a file from the Library and discuss it in chat – PDF and Word now read there too
• The Mode pill is gone: say what you want and chat picks the tool and the settings from your words; the studios keep the full controls

Customer 360 – consent changes on the Activity tab, SMS on the Compliance tab, the blacklist on the page

One person's standing, in one place. The header of every customer says whether they are blacklisted – which tier, which list, since when, by which identity. The Compliance tab shows the present per channel, marketing and service side by side, e-mail, SMS, WhatsApp, push, web and ad audiences, with the reason and the date when a channel is withdrawn. Every consent change – a STOP text, an unsubscribe, a gateway list – now appears on the Activity tab as well, with its own filter.

• SMS and WhatsApp were missing from Compliance because a STOP lives on the phone, not the e-mail – every identity is read now
• A phone-keyed list entry reads as what it is: texts blocked, e-mail untouched
• Blacklist history had been silent on every profile; it lists every entry again, active or lifted, with the identity it was keyed on
• The Customers list carries the blacklist badge beside the name

ProfiSMS blacklist – the gateway's opt-outs, kept in step

If you text through ProfiSMS, its own blacklist for your login now lives in The AI CMO too. Once a day the list is pulled and kept as a blacklist of its own on the new Blacklist tab – numbers on it stop receiving marketing texts, service texts and e-mail are untouched – and a STOP that reaches The AI CMO is added at ProfiSMS at once, so texts sent through the same login from other tools respect it as well.

• Blacklist is now its own tab in Customer Intelligence – every list, the suppressed-send evidence and the gateway feeds in one place, no longer inside Audiences
• One list per brand, "Marketing opt-out feed (ProfiSMS · your login)", with the count and every entry
• A number your team removes at ProfiSMS is lifted on the next pull; a START from the customer removes it at once – unless they are on a block-all list
• A pull that looks wrong – an empty or much shorter file – lifts nobody and says so on the SMS sending card, which also has Pull now

Meta Ads Strategist – the audit your Facebook and Instagram ads never had

The Google Ads Strategist has a twin. The Meta Ads Strategist reads your Meta ad account live – campaigns, ad sets, ads, placements, age and gender, devices, the daily spend – runs a deterministic engine over it and writes an agency-grade audit: an executive summary with the money line, a grade per axis (structure, targeting, creative, measurement, pacing), prioritised recommendations with conservative, balanced and aggressive variants, and the estimated impact. An axis the audit could not read is marked unscored, never assumed clean.

• Apply opens chat: the move is validated first, you confirm, then it happens – pausing a campaign, an ad set or an ad, moving a budget – and every action is recorded with what was there before
• Budgets move at most three times up or to a tenth in one step; anything bigger stays yours to do in Ads Manager
• Creative refreshes, audience changes, tracking fixes and landing pages stay human steps – the audit says why
• "How are my Facebook ads doing?" in chat reads the account live without an audit; the AI team's paid-media specialist uses the same tools

Petře, not Petr – Czech name forms in every message

Czech and Slovak brands can now address customers the way people actually speak. New personalisation tokens give the vocative (Petr → Petře, Jana → Jano, Novák → Nováku), the familiar form (Jan → Honza, Josef → Pepa, Kateřina → Katka) and the customer's gender for pane/paní and dostal/dostala – computed by rule at send time, so the list grows by the name, never by the customer. They work in Email Studio, journeys, SMS, WhatsApp and chat drafts, and the recipient preview shows exactly what will be sent.

• Brand profile → Market: choose how you address customers (formal or familiar) and the greeting {{p.salutation}} renders – empty uses the language's own default
• A name the rules are not sure about is used as written and lands under Names to look at; settle it once and it holds for every message
• Female surnames and adjectival male surnames never change; the spelling the customer gave is the one that is declined

Carousel Builder follows a colour ask

"Bright pastels, not our brand colours" used to be ignored: the carousel had no channel for colours and painted the brand theme whatever the brief said. Now a colour ask in the brief – a mood, a palette, hex values – repaints every slide: a cover colour, backgrounds cycled across the cards, an accent, with the type colour chosen for contrast. Without a colour ask nothing changes and your brand palette applies as before.

Connections, tidied up

A handful of fixes on the Connections page and the Website SDK, prompted by customers setting up their first brands.

• A Website SDK setup can be shared with another of your business profiles – a test brand no longer needs a second implementation; events stay with the brand whose key sent them
• The Connections list shows the brand's connections, whoever on the team connected them, and a connection made from the header's brand is filed under that brand
• The Support inbox card shows Connected and appears in the Connected strip like every other Send channel
• When your sending domain is already registered on another Resend account, the setup says so and offers the bring-your-own-provider door instead of failing silently

Helpdesk: one inbox for the brand's whole team

The brand's seats now share the Helpdesk inbox – conversations, replies and the customer records behind them belong to the brand, not to whoever connected the mailbox. Replies can leave through your own SMTP, so they arrive from the support address customers already know.

Did they get it? Delivery status on every email and SMS

Every send now carries what the provider reported after it left: delivered with the time, bounced or not delivered with the reason, delayed while the provider keeps trying, or sent with no receipt yet. You see it under each send step in a journey's Activity panel and on the person's timeline under Customers. SMS receipts through ProfiSMS are requested with every message and need no setup; email providers report through the webhook on the sending card.

Your logo on the office wall

The strategy room in the virtual office now carries your brand's logo on a framed wall sign – the one you uploaded under brand settings, at its own proportions and colours, on a dark face when the mark is light. Change or remove the logo and the sign follows the next time you open the office; without one, the wall keeps its original lettering.

Hand a project task to an AI teammate

A task's assignee list now includes your AI team. Pick a specialist – or the new Companion, hired for exactly this – and it plans the fewest steps to deliver that task, does the work, and brings the result back to the task's record for you to approve. The task moves to Done when the work is done.

• A goal that needs one more detail from you now asks on the goal page and waits – it used to skip the question
• Fresh ideas and new goals respect the same rails as the rest of the team: the master switch, your plan, and your budgets
• Agents → Office gains a Ledger: every action the team took, who took it, and the trust and budget it ran under – with a CSV export
• Teammates on a brand now see its AI team, its goals and its review queue, and can approve, reject and steer – the brand decides, not the row's author

Document carousels – slides and caption, delivered as the document

Ask Chat for a carousel and it arrives as one: a cover, one idea per slide, a closing ask, rendered as a square-page .pptx your LinkedIn document post takes, with the caption that carries it. The slides show as a filmstrip in the chat and in the library, open in the deck viewer, and Schedule on LinkedIn hands document and caption to the scheduler.

• A carousel task in a project now opens the Carousel Builder, not the post tool – the post tool writes one post's text and never slides
• LinkedIn publishing attaches a document as a document post and several images as a multi-image post; it used to keep only the first image

A virtual office – watch your AI team at work, in 3D

Autonomous has a new tab. Every specialist you employ, and every goal running on its own, is a colleague in a small office: at a desk in Strategy, Creative, Growth or Analytics while on a task, in the lounge between assignments. What each one is doing right now is written next to them and read from the actual run – nothing is staged.

• Working means the job is executing at this moment; scheduled, planning, in review, paused and setting up each show as exactly that
• Click a colleague to see their current step and open their page; the roster lists everyone, and picking a name brings them into view
• Refreshes every 12 seconds while you watch, on top of the live updates you already had
• Pause the animation, switch day and night, move the camera or go fullscreen – presentation only; nothing here starts, stops or changes the work itself
• Loads only when you open the tab, builds its scene locally and respects reduced-motion settings; a clearly labelled demo shows fictional colleagues if you have no team yet

SMS Studio – texts on the same loop as e-mails

Write a text message, check it, review it with your team, send yourself a test from your own sender, and send it to an audience – on one document. Every save is a template in your library, which the journey SMS step already picks from.

• Characters, parts and what is left in this part in display type under the sheet; a split text shows each part as the carrier cuts it, and the phone preview shows one bubble per part
• Personalize inserts fields at the cursor – first name with a fallback, the brand name, any column you imported; Add a link shortens an address into a tracked short link that counts taps; the To line counts the people and the phones, and says how many messages a send bills
• Marketing or service: a service text reaches people who opted out of marketing and leaves from the service sender
• Quality reads it before it ships: placeholders left in, plain http links, opt-out wording you can drop, marketing wording inside a service text, a missing brand name
• Review with status and comments, a [Test] to your phone from the brand's verified sender, and Send that opens a one-step journey – opt-outs, the blacklist, the weekly cap and your budget apply as always

Audiences from who people are, not only what they did

New audience now builds a segment from an attribute – tier, country, balance, a date, any column you imported, on the contact or on the warehouse customer record – on its own or together with an event. The picker is your own data dictionary, so it offers the fields you actually hold.

• Operators that read as a sentence: is, is not, contains, more than, less than, after, before, is set, is not set – text regardless of case, numbers as numbers, dates as dates
• Profile-only audiences refresh every few minutes and before each journey run, like event audiences; rules audiences finally have a Refresh button

The calendar now says who did what – and whether it happened

Open the calendar and every item names its owner: the teammate who did it, the journey that sent it, or the specialist on your autonomous team. A row of faces above the month shows who is active and how they are doing – tap one to see only their work. And a new Week by owner view lays the week out as lanes, one per person, so a Monday stand-up is one screen: who has what this week, what already shipped, what is late.

• Done means something now: every item carries its receipt – done at 3:56 PM by Jana, sent to 4,812 with 41% opened, published on LinkedIn with the link, or failed and why
• A check means it happened, a hollow dot is still to come, red is failed or overdue – and those are filters too, so "what went wrong this month" is one click
• The month line speaks your plan's language – "7 of 122 tasks done · 2 due tomorrow" – and each task says what it is, like "Founder post · LinkedIn", instead of "Other"
• Pick several channels at once – Email and Social together – from the Channel menu
• Big team? The row of owners scrolls sideways, and lanes scroll under the day header. Works on a phone

Make it yours: profile photos, picked avatars, and notifications in your inbox

Settings grew three small things that make the workspace feel like a team. Put a face on your profile – upload a photo or pick one of 24 avatars drawn for The AI CMO – and it shows wherever your name does. And if the bell is not enough, have your notifications land in your email too.

• Upload a photo (JPG, PNG or WebP) or pick an avatar – one click either way, in Settings
• Email me my notifications: mentions, replies and alerts also reach your inbox, batched a few times an hour – the bell keeps showing everything either way
• A Department field on your profile, so bigger teams can tell who does what

Articles keep their history – and always show the final version

Change an article in chat, add its images, or edit it yourself – the copy in your library follows along, so what you rate is what actually went out. And every change keeps the version before it.

• See the history right on the article's page: what changed and when, with one click to bring any earlier version back – and a way back from that too
• No duplicates, no stale drafts – the chat updates the article it already saved for you
• Want it to match your published page exactly? Ask the chat to check the live page in full – it reads the real page, word for word

Identify visitors by your own customer ID, their email, or both

Email is not the identifier – yours is, whichever your data carries. The SDK page and the manual now show identify with your own customer ID, with an email, and with both, and say how to choose: use the identifier the rest of your data is keyed on, so website behaviour joins the same customer record.

• SDK 1.25: when you pass both – identify('player-48213', { email }) – both reach the warehouse and resolve to one person; before, the email silently dropped the ID
• The inbox API and the customers API examples show the customer-ID form alongside email
• A conversion now carries your own fields first-party – the channel a signup came through, the plan, the market – up to 20, stored on the event in your warehouse
• Those fields drive the platform: audiences can be "signed up where channel is affiliate", journey event triggers and waits can require a value, the customer timeline shows them, and messages can use {{p.channel}}

ProfiSMS joins the SMS providers

Czech and Slovak teams can now text customers through their own ProfiSMS account. Pick it in SMS sending, enter your ProfiSMS username and password, and an optional sender ID – we verify the login against your account without sending anything, or send a real test text if you give us a number.

• Same opt-out handling as every other provider: STOP and UNSUBSCRIBE suppress the number on our side, START re-subscribes
• Inbound texts arrive through the ProfiSMS callback URL – the card shows the exact address to paste into the Extranet
• Your ProfiSMS credit, your rates; the password is encrypted at rest and only ever hashed per request

Landing pages that already look like your site

Describe the offer in plain language and a full page builds itself in front of you, section by section: hero, proof, offer, FAQ, lead form, closing call to action. Before it writes a word it reads your own website and takes its fonts, colours, button shape and logo, so what arrives already looks like yours instead of needing an afternoon of styling. It also reads your homepage and up to two of your own pages that carry the offer, plus one search of the web, so the copy argues from what you actually sell. Then you talk to the page: say what should change and it comes back rewritten, or click one section on the canvas and change only that.

• It carries over your fonts, colours, button shape, light or dark theme and your real logo, but not your layout – and it needs a website address on the brand profile to have anything to read
• Up to three images per page, drawn for the brand and placed in the spot the layout reserved for them, each one billed as an image generation like anywhere else in the platform
• Prices, guarantees and testimonials appear only if they came from you or from what the research found, and the copy is still machine-written, so read it before you point paid traffic at it

One audience now holds your whole customer base

Audiences used to stop at 20,000 people, and a segment saved from the churn or RFM lists stopped at 5,000 – numbers that were fine for a short list and quietly wrong for anyone with a real book of customers. One audience now holds up to 500,000 people, a saved RFM or churn segment up to 250,000, and a push to Meta or Google carries up to 500,000. If a refresh matches more people than the ceiling holds, the card says so rather than looking complete, and the push tells you how many stayed behind.

• These are safety ceilings rather than something a plan buys – ask and we raise them for your account
• The refresh reads your customers in pages now, so a large segment rebuilds fully instead of stopping partway through
• Truncation is always stated on screen: an audience that met its ceiling never pretends to be the whole list

Read your customer base back out, over the API

Your data could come in through six doors and leave only as messages. Now a key with the data:read scope can read it back out: a page of profiles for a brand, one person by e-mail or your own ID, everyone in a saved segment, or everything that changed since a timestamp – as JSON, or as a CSV file. That last one is the sync loop: run it on a schedule and another system of yours stays in step with what the platform knows.

• A segment of any size comes out in full – keep passing back the cursor you were given
• Erased people are never returned: an export path must not resurrect someone who asked to be forgotten
• Scope it deliberately – data:read is separate from the scopes that let a key create or send anything

A customer's record, without their e-mail in the address bar

Opening a customer used to put their e-mail address in the URL – which means in your browser history, in your bookmarks, and in the referer header of anything that page loads. Customer records are now addressed by an internal reference instead: the page looks the same, the address gives nothing away, and customers with no e-mail on file open just as well. The Export data button on a record was also handing back an empty file when the identity could not be resolved that way. It now returns the real subject access export, across every rail we hold for that person.

• Old links keep working – a record merged into another still resolves to the surviving one
• The export answers a GDPR access request: profile, identifiers, events, sends, promos, support, surveys, consent and journeys, in one file

We read your live page and tell you if the tag is there

The page where you set up website tracking can now check your site while you watch. Press Check again and we fetch your page exactly as your server sends it, then answer plainly: the tag is there with this site's key, the tag is there but carrying another site's key, the page could not be reached, or the tag is not on the page we read. The status underneath moves through three stages – not seen yet, installed, receiving data – and each stage says what produces the next signal. Page views count as data now, so a site with visitors but no sale yet stops reading as though nothing had ever arrived.

• A snippet copied from another site is caught rather than waved through: we name the key we found and the key this site expects
• We read one page as your server sends it, so a tag added through a tag manager, or one that appears only once the page runs its scripts, honestly reads as not found until a real visitor loads it
• Receiving data can now mean page views alone, so a green status is no longer proof that a purchase or sign-up is wired up

Add a single customer without opening a spreadsheet

There is now a New customer button on Customers, and the same form in your Data hub under CSV import, for the one person a file would be overkill for. Fill in an email or your own customer ID, a name, a phone number and whatever attributes you keep, and the person is on the brand straight away: searchable, with a Customer 360 page, ready to enter a journey. If the identifier you type already belongs to someone on that brand, the two records become one instead of a second copy, and the existing customer survives with its values intact and its blanks filled in from what you typed. Where we can see a likely duplicate, another identity on the same device or the same name twice, a Merge button appears with its reason stated, so you confirm it rather than us deciding.

• Anyone you add by hand is marked a test customer unless you turn that off – test customers stay out of churn scores, segmentation and forecasts and can be hidden from the list, but they still enter journeys and still receive real email and SMS
• A merge cannot be undone, and it stays inside one brand: the same address on two brands is two customers, on purpose
• Consent is never assumed. Marketing email and marketing SMS are tick-boxes, each available only once the matching address or number is filled in, and every one you tick is written to that customer's consent record with its date

Everything your privacy review asks for is now published

When legal or procurement asks for our paperwork, you can send links instead of starting an e-mail thread. There is now a Data Processing Agreement at /dpa, a full Article 28 agreement with annexes, effective 26 August 2026, and it applies as published, so nothing has to be negotiated before it covers you. The privacy policy has been rewritten around a register that sorts every provider we use by what it actually touches: who holds your data, who is infrastructure that never sees it, and which providers of the writing models receive only the brief and the draft, never contact records or event logs. A published cookie policy lists every cookie and stored item with its purpose, duration and recipient, checked against a live scan of our own site rather than written from memory.

• The agreement is in force without a signature, so there is no form to chase – if your legal team wants a countersigned copy or the transfer paperwork, that still comes by e-mail from privacy@theaicmo.com, not from the app
• On our own site nothing in the analytics or marketing groups loads until a visitor chooses, and rejecting is the same single click as accepting; strictly necessary cookies, such as Stripe's on the checkout page, do load without consent, and every one of them is listed
• Your account database and sign-in data sit in the EU (Stockholm), while the warehouse, behavioural analytics and the writing models are EU or US – EU-only processing across all of them is an enterprise option agreed on the order form, not the default

Invite a colleague and they land in the work, not in setup

An invited teammate now arrives straight in your workspace. No brand to set up and no plan to pick: their plan, their credit balance and every paid feature read from the organization's billing, so a seat you are already paying for never looks unpaid. Where the workspace is shared, the Library is scoped to the brand rather than to the person signed in, so campaigns, articles, ads, strategies, press releases, landing pages and surveys made for that brand sit together, signed with the name of whoever made them. A People filter narrows the view to one colleague or a few, and opening someone else's piece opens the real thing in the studio that made it.

• Reading opens up, writing stays with the author: you can open, read and export a colleague's campaign, but Delete and Move appear only on your own work
• Uploaded files, generated images and video, and folders stay personal to whoever made them, and chat threads stay private in every case
• It applies to agencies and to any organization with an invited teammate, and a solo library is unchanged

A finished strategy now fills in the plan for you

Strategy Creator opens on the brand's most recent saved strategy instead of an empty form, and the Setup tab keeps the six most recent one click away. When the strategy is done, Build Marketing Plan carries it into the Marketing Plan Creator with the brief already filled in: goals, channels, monthly spend, duration, industry, audience and brand voice. An organic strategy now hands over a media budget of nothing instead of the default monthly spend the form used to fill in, and the plan is written as an organic plan, in hours and owned channels.

• The hand-off fills the brief, it does not write the plan – you still read it over and press Generate, and that spends credits as usual
• It holds the six most recent strategies for the brand you have selected; anything older lives in the Library, and there is no search across brands

Ask for an article, get the pictures too

Chat now offers image slots when it writes you a long-form article. Say yes and it places them where a visual actually earns its space – a hero under the headline, a diagram where the argument gets technical – and writes a brief for each one. The article arrives immediately; you generate the pictures you want, in place, one at a time or all at once. Every image is drawn for the brand you are working on, using its colours, logo and visual style.

• The slots sit in the article where the image goes, not in a list you have to match up by eye
• Anything you choose not to generate is removed on the way out – a slot you skipped never appears in a published post
• Publishing to WordPress, and PDF, Word and HTML exports all carry the finished images, sized to fit the page

Everything we write now knows whose brand it is

Your brand profile reaches every piece of generated copy. Who you sell to, what you actually do, how you differentiate and the voice you use now travel with each request, instead of only your guardrails. The difference shows most in long-form: articles argue for your buyer rather than a generic one, and stop reaching for the register every company in your category uses.

• Applies across Writing Studio, Email Studio, the ad studios and chat – one change, every tool
• The richer your brand profile, the more it has to work with; the audience field does the heaviest lifting

AI CRO tells you what it actually measured

The conversion review was rebuilt to read like a conversion director rather than a checklist. It now knows your industry and audience, can look up how your market positions itself, and sizes every recommendation from your own numbers instead of quoting a lift it cannot know. Where it had no visitor data, it says so plainly rather than presenting a page review as an analysis of your traffic.

• Attention heatmaps are labelled as predictions, because that is what they are – they read the page, not your visitors
• "Receiving data" now means the tracker really has sent us something
• Sites are registered once, on the SDK page, so the same site cannot exist twice under different settings

Only your guardrails are enforced

Industry suggestions are exactly that. They inform how content gets written and they never judge it afterwards – only the rules you added are enforced, which is how it was always meant to work. If a suggestion is one you want held to, there is now a button to make it your own rule, editable and set to your severity.

• With no rules of your own, nothing runs and nothing is flagged
• Suggestions you adopt become yours – reword them, change the severity, turn them off

A new version no longer interrupts you

When we ship an update, you get a small note in the corner instead of a dialog that has to be dismissed by reloading. Reload when it suits you. Writing Studio also stopped suggesting your next sentence while you were still writing it.

• Reloading keeps your place, including the tab and view you were on
• Unsaved brand profile edits survive a reload, a tab change and a brand switch

A loyalty program your customers can actually spend

Loyalty went from a strategy document to a working program. You decide exactly which events earn points – and what spend means in your business, because a sportsbook bet is not a casino deposit and a SaaS invoice is neither. Before you switch anything on, a simulator prices the program against your real history, so you know what last quarter would have cost you. And when you change the rules mid-flight, you choose what happens to the people already in it: keep their earned tier, or re-evaluate everyone under the new math.

• Earning is an explicit allowlist – you pick the event types from your own data, nothing earns by default
• Rewards, balances and redemptions have a public API, so your product can show points and burn them at checkout – a retried redemption never spends twice
• Tiers, expiry, grandfathering and terms & conditions are all yours to set, per brand
• Each brand runs its own pot – points earned on one brand can never be spent on another

Every brand keeps its own books

We went through the platform end to end and made brand separation a hard rule, not a convention. Loyalty balances, churn predictions, lifecycle stages, customer timelines, alerts – each belongs to one business profile, full stop. A customer who exists on two of your brands appears as two separate relationships, each seeing only its own history. If you run brands for clients, this is the guarantee that one client's view can never leak another's customer.

• The alerts bell and Marketing Pulse now show the active brand's alerts only – switch brands, the alerts switch with you
• Customer views, predictions and loyalty are scoped the same way, by default, everywhere
• Deleting a brand now takes its operational records with it instead of quietly promoting them to account-wide

Flash report: the month you picked, honestly told

The headline cards now show totals for the whole period you selected, not just its final week – with rates like ROAS and CPA computed properly from the period's sums. A week with no observed data now reads as a dash instead of a zero, because "we saw nothing" and "you got nothing" are different facts and only one of them deserves a red light. The report also remembers the period you picked when you navigate away and back.

One home for documentation, and API keys that explain themselves

All documentation now lives under Docs – including a proper API manual with real, copyable code examples – and the old legacy API page is gone. On the API keys page, every key now shows exactly what it can call, so granting a permission is an informed decision rather than a guess.

Service messages – receipts and notices that always arrive

Your messages now come in two kinds, and the platform keeps them honest. Marketing is what it always was – campaigns and journeys, governed by unsubscribes, frequency caps and send hours. Service is new: order confirmations, password resets, payout and account notices. A customer who unsubscribed from your marketing still needs to know their payment went through – so service messages reach them, carry no unsubscribe link, and send from their own verified identity so your campaigns' reputation never touches your receipts.

• Set a journey to Service under Settings → Message type – it re-enters for every event and skips send hours, because a receipt can't wait for business hours
• Service email sends from its own domain – set it up in Connections, next to your marketing identity
• Every service send is checked for marketing content – promo language in a receipt is flagged in the run log before a customer or a regulator sees it
• Hard limits still hold: bounced addresses and a blacklist set to "Blocks everything" stop service messages too

Send receipts from your own system, through your CMO

If your app already knows an order shipped or a password was reset, it can now hand the message to the platform to deliver – on email, SMS, WhatsApp, push, or the on-site inbox – with one API call that sends immediately. Idempotency keys mean a retried request never becomes a duplicate receipt, and every send is logged with what kind of message it was, so your compliance story is a query, not a reconstruction.

• One call, five channels – delivered on the spot, not on a schedule
• Grant a key the permission under Settings → API keys – existing keys can't send until you say so
• Suppression and blacklists are enforced per brand, exactly like every other send

Storage is unlimited, and Usage shows everything you have left

Storage limits are gone – generate and keep as much as you like, and we'll simply show you how much you have. The old storage tab in Settings is now Usage: credits, email sends against your monthly allowance, warehouse events against your plan, and storage, all in one view with the same numbers you'll find on the Data page. If a meter is running high, it says so – and says what happens at the top, plainly.

Every connection card now has its verbs

Connected used to be the end of the story on the Connections page – a green pill and no way back in. Now every sending channel keeps a Manage button beside its Connected state (rotate a key, change a from-address, adjust a cap), and every connected platform card can disconnect right from the card, with a second tap to confirm so one stray click never severs anything.

Your specialists now have their own email address

Every autonomous specialist can write to you from their own address – a question when they need your input, a heads-up when a plan is waiting on you, a note when a goal is finished. Reply to the email and they pick up your answer and keep working. You stay in charge of anything that publishes: approvals still happen in the app.

• Each specialist writes from their own address, under their own name
• Reply from your inbox – your answer reaches them and work continues
• Only you can steer: replies from anyone else are ignored
• Turn it on under Autonomous → Office → Notifications → Delivery channel

A calendar for your AI team

Autonomous Mode has a Calendar tab: a month view of everything your specialists did and everything they have scheduled – work published, reports produced, items waiting on you, goal milestones, and upcoming runs. With one specialist the team list is enough; with many, the calendar is how you see the tempo of the whole operation. Filter by specialist to follow one person's month.

Every checkpoint now shows its work

Goal pages used to tick checkpoints off without showing what they produced. Now each completed step shows its actual output – research findings, data pulled from your warehouse rendered as a table, documents with a link to the full result. Goals also show credits spent against their estimate, and cloud-browser research now delivers its findings instead of finishing silently.

Agent reports: everything your team files, in one tab

Analytics has a new Agent reports tab listing every analysis document your tools and autonomous goals produce – performance reviews, KPI dashboards, board decks – each traced to the tool, goal, or schedule that made it. Decks open in a slide viewer and download as presentations. If your team said it built a report, this is where you find it.

Autonomous work actually rings the bell now

Notifications from your AI team – a draft ready for review, content auto-published, a budget warning, a plan waiting for approval, a finished goal – now arrive in the notification bell with a link that takes you straight to the right place. They were being recorded somewhere the bell never looked. The Notify me step in Journeys can also email you a copy of its alert, so results reach you even when you're not in the app.

Autonomous Mode speaks plainly

A full pass over the autonomous surfaces so they read like a colleague, not a console. Statuses are plain language ('Waiting for the scheduled time' instead of system codes), countdowns actually count down, drafts blocked by brand guardrails show exactly which rules they tripped, and checkpoints that will publish somewhere real are marked 'goes live' – no badge means the work stays inside the platform. Campaigns now shows only the work you commissioned; what your specialists start on their own lives under Autonomous, attributed to them. Credit pricing is stated honestly everywhere: tool runs cost their listed price, thinking check-ins 0.5, a batch of specialist ideas 2.

Experiments – A/B tests as decisions

Give a personalization rule two variants (or a holdout) and wrap it in an experiment. The new Experiments page shows each arm's exposed visitors, conversions, revenue per visitor, uplift against the original, and the probability the winner is real – measured on identified visitors, with small samples flagged honestly. Declaring a winner applies it for everyone in one click.

Winners earn their next move

When a piece of content beats its own channel's two-week norm, it now shows up on Home as a move, not a chart – with a one-click Repurpose that opens Chat carrying the winning content and its numbers. Channel-relative on purpose: an email that doubles email's norm outranks a big absolute number elsewhere.

The sidebar tells the loop: make, engage, measure

A new Customers section gathers Audience, Inbox, Surveys, and Promos – the surfaces where you know, talk with, and reward your customers. Analytics moved next to the rest of measurement under Insights, and Setup at the bottom keeps pure infrastructure (Integrations, Data) out of the daily flow. Every path is unchanged, so bookmarks and muscle memory survive.

The Data Manager tells one person's story

Example values across the Customer and Purchase sections now come from one real record instead of a mix of rows – no more two different emails pretending to be the same customer. Demo data got the same care: named demo people with stable external IDs, so customer identification is something you can actually test.

Home shows your latest work, named by what it says

The "Lately from you" strip on Home now mirrors your Library exactly – same items, same order, landing pages included. Creations are titled by their content (the ad's headline, the brief's first line) instead of the tool that made them, and research and campaign cards show a readable digest instead of an empty box.

WhatsApp joins your journeys

The fourth messaging rail, next to email, SMS, and push. Connect your own WhatsApp Business account (Meta Cloud API) once, and journeys can message customers on WhatsApp – using your Meta-approved templates, with variables filled per send and a preview of the message as the chat will show it.

• Bring your own WhatsApp Business account – verified against Meta before saving
• Approved templates with {{variable}} filling – compliant by construction
• Send to the person in the run, or broadcast to an audience
• Opt-outs shared with SMS; blacklist and self-exclusion rules apply before anything sends

Surveys – ask your customers, and your data gets smarter

NPS, satisfaction, and open feedback on a page that carries your brand – logo, colors, and wording inherited from your business profile. Describe what you want to learn and the AI drafts the questions in your brand's voice. Every answer connects back to the person: identified respondents join your Contacts stamped with their score, responses land in your warehouse as survey_completed events, and the results panel joins answers to revenue – so you can see what a promoter is actually worth against a detractor.

• Branded fill pages – logo, accent color, intro, all from your business profile
• Every answer is recorded as it's given – partial and completed responses shown separately
• An identity question (email, username, or customer ID) joins answers to your data even on public shares
• Answers → performance: revenue per person compared across promoters, passives, and detractors
• Journeys trigger on "completes a survey"; survey links in campaign sends are personalized per recipient
• Ask in chat: "create a survey and write an email for it" – your agents create, read results, and link them
• Optional tracking on the fill page: your own SDK site, GA4, or Meta Pixel

Link Building – SEO articles written, published, and earning links, on autopilot

Rankings are earned twice: by what you publish and by who points at it. The new Link Building studio runs both sides as a paid add-on. It researches keywords worth owning for your market, writes genuinely useful long-form articles, publishes them continuously, and earns contextual backlinks from relevant sites as the library grows. Everything lands as receipts, not promises: a publishing calendar with the planned queue, every link listed with its source, status, and domain rating, and articles delivered into your library – publishing straight to your connected CMS when you want them to. You stay in control of the plan: change any queued keyword until its article is written, and give the writer a brief per article. Your autonomous specialists see the program too, so they complement it instead of duplicating it.

• Publishing calendar – what shipped and what's queued, day by day
• Every link earned, with source site, verification status, and domain rating
• Change queued keywords and add writer briefs until the article is written
• Articles land in your library and publish to your CMS – or wait for review

Journeys grow into scenarios – multiple triggers, send windows, and a canvas you can annotate

One journey can now carry several entry points: an event trigger and a schedule in the same scenario, each with its own audience scope, each switchable off on its own. Sends respect per-day send windows – 9–10 and 12–14 on weekdays, quiet weekends – set visually, per journey. Waits got smarter too: hold until a customer does something (or doesn't, and branch on it) or until a specific moment. And the canvas itself now takes notes and section frames with size and color presets, so a complex scenario reads like a document, not a puzzle. The journeys list keeps up at scale – search, filters, and triage sorting when you're running dozens.

• Multiple triggers per scenario, each scoped and individually switchable
• Per-day send windows with multiple intervals – set once, respected everywhere
• Wait for an event with a timeout branch, or wait for a moment
• Notes and section frames make big scenarios readable

Image Center, refreshed

One of our oldest studios caught up with the rest of the product. The editor chrome is cleaner and quieter – grouped controls, one clear Save, a calmer canvas – and generating starts from a proper brief: describe what you want in a sentence, pick a format from named presets (Feed 4:5, Story 9:16, Cinema 21:9…), and the studio expands it into a full creative prompt before generating. You see that prompt, can edit it, and regenerate until it's right. A handful of rough edges went with it, including a bug where vague prompts could come back with no image at all.

• Brief in, agency-grade prompt out – visible and editable before you regenerate
• Named format presets instead of a ratio dropdown
• Cleaner toolbar, dialogs, and canvas – same tools, less noise

Your numbers in your currency – set it once per brand

If your store charges in koruna, euros or pounds, your revenue was always koruna, euros or pounds – the platform just labeled it with a dollar sign. That's fixed. Set Business currency in Settings and every revenue and customer-value figure follows it: Home, Analytics, Customer Intelligence, Marketing Pulse, Flash Reports, the Data Warehouse, win-back audiences, your morning brief. Formatted properly for the currency, not just a swapped symbol – 57 214 Kč, not Kč57,214. Each brand has its own setting, so an agency can run a Czech client and a US client side by side. And closed flash reports keep the currency they were closed with – frozen evidence doesn't rewrite itself when a setting changes.

• Per brand, not per account – every client can charge in its own currency
• Everything that shows business money follows it, including exports
• Closed report packs stay exactly as closed – evidence never re-labels itself
• Your subscription and credits stay in USD – this is about your data, not our pricing

Lifecycle Movement answers the two questions that matter: is this normal, and is it covered?

"$84,276 slipped" means nothing on its own – maybe you also recovered more than that. Movement now leads with net value drift: what slipped, what recovered, and how that compares to the period before, so you know whether this week is a crisis or a Tuesday. Below it, automation coverage shows what share of the slipping value has a live journey running against it – and where nothing is listening, one tap builds the play with the right trigger already set. The whole page is built for operators: momentum in money rather than headcounts, the matrix shows value moving with each cell's share, and every ribbon in the flow chart is clickable – see exactly who moved, what they're worth, and act on them.

• Net drift with a baseline – slipped vs recovered vs the prior period
• Automation coverage – how much slipping value has a live play, and the gaps
• One-tap "Build the play" seeds the journey with the right lifecycle trigger
• Click any ribbon or cell to open the flow – the people, the value, the actions

Win-back picks up where you left off – and drafts every touchpoint

Your latest generated campaign now survives a refresh: come back any time and it's waiting, with its numbers intact. Every touchpoint card gained its own draft button that knows the channel – an SMS touchpoint drafts an SMS, a call touchpoint drafts a call script – each seeded into Chat with the full brief: angle, offer, timing, audience and its real size. The audience card's numbers now reconcile to the person with the churn tab and Home, the campaign exports to PowerPoint, and building the journey states plainly what it does: the structure and timing – the messages themselves are yours to draft, per touchpoint or all at once.

• Refresh-proof – the latest campaign restores automatically, per brand
• Channel-aware drafting on every touchpoint – email, SMS, push, call scripts
• PowerPoint export alongside Word and text
• Audience numbers reconcile exactly with Customer Intelligence and Home

Chat can now find what it doesn't have a URL for

Ask "who are the top competitors in my market?" and your CMO can actually go look – it searches in your market's country and language, not generic US results, then reads the pages that matter. Name a specific site and it still browses it directly, as before; search is only for discovering what neither of you has a link to. Long research runs got sturdier too: when a browsing session brings back more than fits in the conversation, the full findings are saved to your library automatically and the conversation works from the essentials – nothing gets lost, and your CMO can pull the details back in whenever it needs them.

• Market-aware search – a Czech business gets Czech results, not US ones
• "Go to their site and summarise it" still browses directly – search never gets in the way
• Long research findings save to your library automatically – interruption-proof

Write it here, publish it there – your CMS is now one click away

Connect your CMS – Webflow, WordPress, Ghost, Contentful, Notion, Medium and more – and it simply appears in the Writing Studio export menu as "Push to Webflow". Pick the collection or blog it belongs to and the article lands there as a draft, formatted and ready for your final review. It works in chat too: tell your CMO "publish this to Webflow" and it asks which collection – in plain words, like a colleague would – then confirms when the draft is live. Everything arrives as a draft on purpose: nothing goes public without you.

• Any connected CMS shows up in Writing Studio's Export menu automatically
• Chat understands "push this article to my blog" – and asks before it guesses
• Always a draft first – you publish from your CMS when you're ready
• Connected-app answers in chat now read like a colleague, never raw data

Your language, everywhere – not just in chat

Set your output language once in Settings and every word the platform writes follows it: strategies, emails and sequences, ads, blog posts, your morning brief, autonomous drafts, journey content – even the headline written inside a generated image and the script an AI actor speaks in your videos. Czech, German, French, Spanish, Polish and more, with proper native spelling and diacritics, not machine-translated English. If you work in one market and write in another, the setting follows the work, not the interface.

• Every generator honors it – tools, briefs, autonomous mode, journeys, chat
• Text rendered inside images and spoken in video scripts follows it too
• Native-level writing with correct diacritics – háčky a čárky included

Analytics you can actually interrogate – any dates, any device, one truth

Revenue analytics gained a real calendar: pick any date range – "June 1 to June 10" is now an answerable question – and every number, delta and channel breakdown recalculates against the equal period before it. Web analytics split by device, so you can see how mobile visitors behave differently from desktop across traffic, engagement, campaigns and site speed. And underneath it all, one calendar of truth: the Dashboard, Analytics and Flash Report now count the same dollars for the same days, to the cent – with a date picker on the Dashboard too.

• Calendar range picker on Revenue, Web and the Dashboard
• Desktop / mobile / tablet split across every web report
• The same question now gets the same number on every page
• The purchases ledger says exactly what its search filters – just the table

Your website becomes a marketing signal – no developer required

The no-code event capture grew up. Point an event at a click, a form submit, or simply a page visit – "reached /thank-you" now counts, even on single-page apps – and pull the order value straight out of the page, like a cart total. Then the part that makes it matter: your website activity now feeds your marketing. Journeys can start when an identified customer is active on your site, and Audiences can hold a living "Active on site – last 7 days" segment, built for the two plays that pay: catching hot visitors while they're browsing, and noticing loyal customers before they drift away. Every plan includes a generous event volume – up to a million on Starter – with a quiet meter in your Data hub so you always know where you stand.

• Capture on click, form submit, or page visit – SPAs included, zero code
• Read values off the page itself – a cart total becomes revenue data
• New journey trigger: "someone is active on your website"
• New segment preset: "Active on site" – target the warm, rest the dormant

Your work finishes even when you don't wait for it

Start something – an image, a video, a landing page, an email, a strategy – and walk away. Close the tab, switch tools, shut the laptop: the work finishes on our servers, lands in your Library, and writes itself back into the conversation. Come back and the chat shows the finished result, not a spinner; stay on the page and it appears live the moment it's done. Landing pages arrive as editable Builder drafts and Email Studio generations save themselves as draft projects, so nothing you generate can be lost by leaving. And one more piece of fairness that was overdue: if a generation fails, the credits come back – automatically, every time.

• Leave any time – generation continues server-side and lands in your Library
• The chat updates itself with the finished result, image and video included
• Landing pages and Email Studio generations persist as editable drafts
• Failed generations refund their credits automatically

The Chrome extension becomes a marketer's sidekick

The extension used to be a capture tool with a chat in a drawer. Now it's your CMO looking over your shoulder, wherever you work. Open the sidebar on any page and it knows where you are – a competitor's site, an ad library, your own ads manager, an article – and offers the three actions that fit, from "What are they testing?" to "Propose one optimization." Right-click in any text field on the web to draft, rewrite, or tighten in your brand voice – LinkedIn composer, your CMS, anywhere. Select what you're about to publish and run a preflight: a ten-second verdict against your brand voice and your guardrails before it goes out. Save a competitor's ad and one click makes your version. And when your autonomous team has drafts waiting, they appear right in the sidebar – approve without leaving the page you're on. Chat responses stream live with a running status line, no more doubled answers.

• Page-aware sidebar – the right three actions for whatever you're looking at
• Write anywhere in your brand voice: draft, rewrite, tighten in any text field
• Preflight before publishing – brand voice + guardrails check in seconds
• Approve autonomous drafts from the sidebar; clip an ad, make your version

“Build a presentation from this” – and it's a real PowerPoint

Ask the chat for a presentation and you get an actual .pptx file – not an outline to paste somewhere. Say it about anything in the conversation: the strategy you just built, a campaign's numbers, research results. Your AI CMO writes the deck the way a strategy team would – an assertive takeaway per slide, your real numbers pulled forward, a closing slide with the next steps – and hands you the download. It arrives in your brand by default: your colors, your logo on the title slide, your font. Want it plain? Say “make it neutral.” Every deck is saved to your library, and regenerating with a different angle or audience is one more sentence. 10 credits per deck.

• A downloadable .pptx, built from what's in the conversation
• Deck-team writing: one takeaway per slide, real numbers, clear next steps
• Your brand by default – colors, logo, font; neutral on request
• Saved to the library; regenerate with a new angle any time

A campaign chat never forgets it's a campaign

However you come back to the chat – sidebar, bookmark, plain link – a coached-campaign conversation now returns with its campaign context intact: the coaching banner, the plan, the step tracking. Before, returning through the plain chat link reloaded the same conversation as a normal chat, and your AI CMO quietly lost sight of the plan it was coaching you through. The conversation itself now carries its campaign identity. Starting a new chat still gives you a clean slate, and switching between conversations in history can't mix one campaign's context into another.

• Come back any way you like – the campaign context comes back with you
• New chat is a genuinely fresh chat; your campaign's conversation stays one click away
• Switching chats in history keeps each campaign's context to itself

Your work in flight, one glance closer – Home, Calendar, tasks

Three small moves that shorten the distance between seeing work and doing it. In-progress projects on Home went from three quiet cards to a working ledger: each row shows how far along the project is (3 of 8 done, with a progress bar), the actual next task by due date, and the deadline – overdue first, so the right thing nags. On the Calendar, clicking a project task now opens that task's detail inside its project instead of dropping you at the project door. And the task card itself got the full treatment: wider, the title in editorial type, the description as open text instead of a cramped box, and the details in a calm two-column layout.

• Home: progress fraction + the next due task for every project in flight
• Calendar: a due task click-through lands on the task, not just the project
• Task detail: more room, clearer hierarchy, description that reads like a brief

Campaign coaching keeps pace with you

When your AI CMO coaches you through a campaign plan, finishing a step now feels like finishing a step. The moment the step's work lands in the chat – a scrape, keyword research, an analytics pull, a draft – a quiet card appears right beneath it: mark it done, and the CMO picks up the next step in the same breath. No more opening the full plan to tick a box. The CMO also keeps perfect count now – marking a step done moves it to the very next one, never past it – and a fast click on the card can't outrun the chat anymore: the follow-through always arrives.

• Mark a step done right under the work – the plan sheet is for overview, not chores
• The CMO continues with the next step, exactly the one the plan says
• Click the moment the card appears – the next step always follows

Journeys, Campaigns, Autonomous and Strategy Tracker – one calm design language

Four of the platform's busiest pages now read like the rest of the house. Journeys went from fifty icons to one calm page – describe the journey you want in a sentence, or start from a template that tells you when it runs and what it ships. Campaigns opens with the number that matters: how many decisions are waiting on you. Autonomous puts the working surface first – what the team shipped, what needs your eyes – in the same quiet ledgers you know from the rest of the platform. Strategy Tracker reads as a pipeline now: which strategies are in execution, what each one has shipped, all in one register. Status everywhere is a small colored dot and a word – no badge soup, no icon noise, charts keep their color where color means something.

• Journeys: describe it in a sentence, or pick a template that says when it runs
• Campaigns: the decisions waiting on you, front and center
• Autonomous: drafts and shipped work in quiet, readable ledgers
• Strategy Tracker: your strategies as a pipeline, execution at a glance

Strategy → plan → project – one thread from thinking to doing

The three levels of your marketing now hand off to each other. Finish a strategy and take it two ways: straight into a project (tactics and content ideas become tasks with deadlines), or into the Marketing Plan Creator – where the plan doesn't just reference your strategy, it executes it. Every tactic gets scheduled at the strategy's phasing, quick wins land in week one, and the duration follows the horizon you chose – a 30-day sprint becomes a 30-day daily plan, a year becomes a year. Your goals, channels and budget carry over so you never retype what the strategy already decided. Then one click turns the plan into a project: every activity becomes a dated task, weekly milestones included, still linked back to the strategy it all came from. The chat knows the whole chain too – ask it to build a plan from your latest strategy and it pulls the real thing from your library.

• Build a marketing plan FROM a completed strategy – it schedules the tactics, not new ideas
• Duration, goals, channels and budget inherit from the strategy automatically
• Plan → project in one click – dated tasks, weekly milestones, all linked back
• Works in chat too: "build a plan from my latest strategy"

Studios search understands what you're trying to do

Search in the Studios used to match tool names only – type "SEO" and Keyword Research wouldn't show up, because the word isn't in its name. Now search speaks the same vocabulary as the tools themselves: what they're for, not just what they're called. "SEO" brings up Keyword Research, Backlinks Analyzer and Competitor Spy alongside the SEO writers; "churn" finds Customer Win-Back; "win back" ranks the right tool first. The quick filters (Write, Plan, Launch…) use the same understanding, so tools land in the right group even when their card never said the magic word.

• Search by intent – "SEO", "churn", "landing" – not just by tool name
• Results ranked by how well they match, best first
• The Write / Plan / Launch filters got the same brain

Home reads like the rest of the house – and greets you properly

Your morning brief now opens in the same editorial voice as the rest of the platform: the lead sets in display type with the numbers and names that matter carrying the ink, this week's work reads as a quiet stat strip instead of boxed tiles, and every section speaks the same typographic language. The greeting got smarter too – it uses the name from your profile, first name only, and if it doesn't genuinely know your name it simply doesn't guess one.

• The brief leads in the house editorial style – key numbers and names stand out
• This week's shipped / rated / drafts as a clean stat strip
• Greets you by your actual first name from your profile

Customer Win-Back – built on your real churned customers

Win-Back used to need a HubSpot connection to find who left. Now it reads your actual customers straight from your data – the built-in warehouse or a dedicated ClickHouse cluster, whichever you're on – and builds the campaign around the people genuinely slipping away. Choose an inactivity window (30, 60, 90 or 180 days) or an at-risk lifecycle segment, and it shows you exactly how many customers that is, the revenue on the line, and their average lifetime value – then writes a win-back campaign sized to those real numbers instead of guesses. When you're happy, one click turns it into a multi-touch email Journey – a drip that sends, waits, and follows up on its own – handed to you as a draft to review and publish. HubSpot still works if you use it, it's just no longer the thing holding the tool up.

• Your churned audience comes from your own data, not a required integration
• Works with the built-in warehouse and a dedicated ClickHouse cluster alike
• Real numbers up front – who's at risk, the revenue on the line, average lifetime value
• Turns the campaign into a multi-touch email Journey you review and publish

Projects, rebuilt – and your chat finally knows them

Projects got the same craft as the rest of the platform. The board reads at a glance now: each task is a clean card with the one thing that matters – what's overdue, what's blocked, what's high priority – instead of a row of coloured labels fighting for attention. The month calendar, the by-tactic view and the list all speak the same quiet language, so nothing shouts unless it needs you. And the header opens with the numbers you actually track: done, in motion, overdue, and the date it's due. The bigger change is that chat is no longer blind to any of it. Ask what projects you have, what's due this week, or what's overdue, and it answers – with links straight to the project. Turn a strategy into a project and it's part of the same conversation from then on.

• The board rebuilt – one clean card per task, no wall of coloured labels
• Overdue and blocked read at a glance; priority only speaks when it's high
• Board, calendar, by-tactic and list all redesigned to match
• Ask chat what's on your plate – it names your projects and links you there

Your project deadlines now live on the calendar

The marketing calendar already showed everything you shipped and everything scheduled to go out. Now your project task deadlines sit right alongside them, so one view answers what's actually due and when – the email going out Thursday and the task due Friday, on the same grid. They come in under a Projects filter you can switch on or off like any channel, so the calendar can be pure send-schedule one moment and your full workload the next. A deadline behaves like a deadline, not a broadcast: it never inflates your shipped and scheduled counts, it reads red the moment it's overdue, and clicking it drops you straight into the project it belongs to.

• Task due dates appear beside your sends and scheduled posts
• A Projects filter toggles the whole task layer on or off
• Overdue tasks turn red instead of quietly fading
• Click any task to jump straight to its project

Backlinks Analyzer – who links to you, and who links to them

A new studio that reads the live link graph for any domain – yours or a competitor's. One run shows how the web ranks the domain's authority, how many backlinks and referring domains point at it, what anchor texts the web uses when it links there, and the strongest pages pointing in – each flagged dofollow or not. Point it at a competitor and you see exactly where their authority comes from, which is a link-building shortlist you didn't have this morning. The chat can run it mid-conversation too – ask who links to a domain and the numbers arrive in the thread.

• Authority rank plus live backlink and referring-domain counts
• The strongest pages linking in, with dofollow flagged on each
• Anchor texts – what the web calls you when it links to you
• Export the full list as CSV, up to 1,000 rows per run – 2 credits

Strategy Creator – from a 30-day sprint to a full year

Strategies now come sized to your horizon: 30 days, 90 days, 6 months, or 12 months. The horizon changes the shape of the work, not just its length – a 30-day strategy arrives as a week-by-week sprint where everything is startable within days, while a 12-month strategy is phased by quarter, accounts for your seasonality, and never leaves a month empty. Turn one into a project and the timeline follows honestly – a sprint creates a four-week project, a year creates a year. The studio itself was rebuilt around how a strategist actually works: your brief reads as a document written from your profile, the decisions are laid out as calls to make, and a launch bar tells you the moment it's ready.

• Four horizons – 30 days, 90 days, 6 months, 12 months
• Year-long strategies phase by quarter and build on each quarter's results
• Project timelines and end dates follow the horizon you chose
• The brief, the calls, the launch – the whole studio rebuilt

Chat answers in seconds – and keeps its reasoning

Chat used to sit silent while it worked and then deliver everything at once – on longer requests that silence could stretch past half a minute. We rebuilt how a turn starts: the first words now land within a few seconds, you watch the thinking stream live while it works, and when the reply finishes the reasoning doesn't vanish – every message keeps a quiet Reasoning toggle you can open any time to see how it got there. One more correction while we were in the room: asking for an ad now gets you one strong ad. Variations are yours to ask for, not ours to assume.

• First words in seconds – no more silent wait before the answer
• Thinking streams live while it works
• A Reasoning toggle stays on every finished reply
• One deliverable by default – say the number when you want variations

Six studios, rebuilt end to end

Keyword Research, ROI Calculator, Loyalty Strategy, Marketing Plan, Referral Program, and Customer Win-Back all moved to the same design language as the rest of the platform – and the same logic. Instead of a wall of form fields, each studio opens with your brief already written from your business profile, reading like a document with an edit toggle rather than a questionnaire. The decisions you actually own – program type, goals, channels, budget – are laid out as clear calls. A launch bar tells you what's missing or that you're ready, and when the work is generated, the export screen leads with what to do next instead of a grid of download buttons.

• Your brief arrives pre-written from your profile – edit it, don't retype it
• Decisions as clear choices, not checkbox walls
• A launch bar that names what's missing before you generate
• Results and exports redesigned around the next step, not the download

Plan changes without leaving the app

Downgrading or cancelling no longer sends you to an external billing portal. Choose a lower plan and the change is scheduled for the end of your current period – you keep your plan, credits and features until the day you've already paid for runs out, and the confirmation tells you exactly which date that is. Cancelling spells out precisely what you lose before you confirm, and if you change your mind before the date lands, resuming takes one click. The plan cards themselves got honest, too: each one now knows whether it's an upgrade or a downgrade from where you stand.

• Downgrades schedule for period end – you keep what you paid for
• Cancellation shows exactly what you lose before you confirm
• Changed your mind – resume in one click before the date lands
• Every plan card knows if it's an upgrade or a downgrade for you

The dedicated Data Warehouse – now one click to activate

The CMO plan includes a dedicated ClickHouse warehouse, and switching it on is now yours to do: one click in Analytics or the Admin Portal assigns your cluster and unlocks the full catalog of data sources. From that moment your events stream into infrastructure built for scale – billions of events, with the same dashboards, Customer Intelligence and segments reading from it. It's a team affair by design: the workspace owner or an admin activates and configures, managers can watch and trigger syncs, and every feed belongs to a business profile – an agency runs a separate set of sources per brand, and each brand's dashboards see exactly their own data.

• Activate in one click – owner or admin, on the plan that includes it
• Shopify orders and Google Ads & Meta ad spend sync in daily on their own
• Every source is tied to a business profile – per-brand feeds, per-brand numbers
• Setup of database connections (Snowflake, Postgres, BigQuery…) is included with the plan

Data Sources – a control room for every feed

The Admin Portal grew a page that answers one question at a glance: what feeds my warehouse, and how fresh is it? Every source shows its live sync status – last run, rows moved, per brand. Adding one starts with the business profile it feeds, then a catalog grouped by how each source works: webhook and CSV work the moment you add them, Shopify and the ad platforms run through the integrations you've already connected, and databases are connected by our team. Sync now means sync now – pressing it runs the real pipeline, and the freshness you see is the freshness you have. Configuration lives here, away from day-to-day dashboards, where owner and admin permissions guard it.

• Live status per feed – last sync, rows, per business profile
• A source catalog grouped by how each one works, not a guessing game
• Sync now runs the actual pipeline – freshness never lies
• Owners and admins configure, managers monitor – enforced, not implied

Stream events by webhook – on every plan

Your systems can now push events straight into your warehouse. Add the Webhook source and you get a live endpoint with its own ingest key – POST a purchase, a signup, any event from your backend, your store, or any tool that can send JSON, and it lands in seconds. One event or batches of a thousand, with copy-paste examples for cURL, JavaScript and Python. Send the same event twice and it counts once – retries and replays are safe by design. It feeds the built-in warehouse on every plan and the dedicated cluster wherever one is active, and everything downstream updates on its own: revenue, lifecycle, segments, the daily brief.

• A live endpoint + ingest key per business profile
• Single events or batches up to 1,000 – landing in seconds
• Repeats never double-count – safe retries built in
• Ad-spend CSVs now flow into the dedicated warehouse too

The Admin Portal, remade for teams

The workspace admin now wears the same design as the rest of the product and tells the whole story of your team's work. The dashboard shows your credit pool, this month's content, and a live feed of what your people created and spent. Usage breaks credits and content down per member; the audit log is a readable trail of who did what, when. Billing reads from the one wallet your whole workspace shares, shows every current plan with honest seat and brand numbers, and keeps plan changes where they belong – with the owner. Managers get a seat at the table too: team, profiles and monitoring, without the keys to billing.

• One credit wallet for the whole workspace – the numbers everyone sees agree
• Per-member usage and a real activity trail
• Billing with every current plan and seat counts that match what's enforced
• Owner, admin and manager roles with the rights you'd expect

Your website is now a channel

Journeys can publish straight to your own site. Drop a Website message step into any journey and visitors see it on your pages – a corner card, a banner, or a modal, shown to everyone or to one specific person a journey is working on. It comes with a real designer, not four fields: image placements (on top, beside the text, or behind it), your brand colors and font one click away, buttons that match your site's design system, widths, alignment, a 'No thanks' line if you want one. And the preview isn't a mock – it's your actual homepage, rendered by the same code that runs on your site, switchable between desktop, tablet, and phone. You decide when it appears: immediately, after a delay, once the visitor scrolls deep enough, or as they're about to leave. Then set a goal – a purchase, a conversion, a form submitted – and it's measured against your data warehouse: people who saw the message and then did the thing, counted honestly, with revenue.

• Corner card, banner, or modal – site-wide or targeted to one person
• A full designer: images, brand colors and font, buttons in your site's style – previewed on your real homepage across desktop, tablet, and phone
• Show it immediately, after a delay, at scroll depth, or on exit intent
• Goals measured from your data warehouse – purchases, conversions, form submits, with revenue and an attribution window you choose
• Impressions, clicks, and dismissals land in your warehouse first-party – ad-blockers don't blind it

Site SDK – one light script, now SPA-native and ad-blocker-proof where it counts

The script you install once on your site grew up. Single-page apps are tracked properly – route changes register as pageviews with fresh page context, no extra wiring. Sites with a consent banner get a real consent mode: nothing loads or tracks until the visitor opts in, and a no persists across visits. And the events your revenue math depends on – conversions, form submits, identify – now reach The AI CMO first-party, so they land in your warehouse even when browser extensions block third-party analytics. Website messages ride the same first-party path, so the new channel works for every visitor, not just the ones without an ad-blocker.

• SPA route changes tracked automatically – no extra code
• Consent mode: hold everything until opt-in, respect opt-out across visits
• Conversions, form submits, and identify mirrored first-party into your warehouse
• Message delivery and install detection no longer depend on third-party scripts

Calendar – everything shipped and scheduled, in one place

There's a new page next to Home and Dashboard: one calendar for everything your marketing did, is doing, and will do. Journey sends land on it with receipts – sent, opened, clicked – right on the day they went out. Scheduled posts, planned work, and everything the autonomous agent ships appear with dates, and a 'Shipping next' rail answers the question every founder asks: is anything actually going out? Filters keep a busy account readable – by source (planned, manual, journeys, autonomous) and by channel – and every item opens its detail: a journey send opens that journey, autonomous work opens the piece in your library. Empty days aren't dead ends either: any open day offers 'Plan this day', which hands the date straight to your CMO chat.

• Past with results: sends carry sent · opened · clicked on the calendar itself
• Month and week views with filters by source and channel
• 'Shipping next' – the next items going out, with relative time
• Every item opens its detail – the journey, the library piece, the scheduler
• Empty days offer 'Plan this day' – the date goes straight to the CMO chat

Journeys – Webhook step: call any API from a run

Journeys can now talk to the rest of your stack. Drop a Webhook step anywhere in a journey and it calls any URL when the run gets there – pick the method, add an optional auth header, and shape the body with variables like the contact's email or, most usefully, whatever the previous step produced. That means a journey can research a topic or write a message, then POST the result to Zapier, Make, Slack, or your own backend. Leave the body empty and a clean JSON envelope goes out, so wiring a catcher needs nothing but the URL. Webhooks respect the same rules as every send: they wait for approval in draft mode, skip in test runs, and stay silent for holdout customers so your uplift numbers stay honest.

• Any method, any URL, optional Authorization header
• Body templates with {{contact.email}}, {{previous.output}}, and more
• Empty body sends a ready-to-catch JSON envelope – Zapier wiring in one paste
• Inherits approval, test-mode, and holdout behaviour from the journey engine
• Private and internal hosts are blocked, with timeouts and retries built in

Journeys – automations that arrive built, and prove what they earn

Journeys is the automation layer of The AI CMO, now in beta for everyone. Draw a welcome series, a win-back, or a launch sequence once and it runs day and night – triggered by what customers actually do: joining a segment, submitting a form, clicking a tracked link, or simply on the schedule you set. The difference from every flow builder you've used: the journey arrives built and written. The emails inside come drafted in your voice and your brand template, SMS steps send through your own number with opt-outs handled, and steps can think mid-flow – research a topic, read a web page, write or design content – so an automation can produce, not just route. And when you want to know what a journey is really worth, switch on a holdout group: some customers don't get it, and the revenue difference is what it caused. Not attribution credit – proof.

• Welcome, win-back, and launch journeys arrive drawn and written – you approve, they run
• Triggers on behavior: segment entry, form submits, link clicks – or on a schedule
• Conditions, A/B splits, and goals on every journey
• Steps that think: research a topic, read a page, generate content mid-flow
• Holdout groups measure the revenue a journey actually caused, not just touched

Campaigns – tell it the outcome, watch the work ship

Campaigns is the new home for the outcomes your AI CMO is running. Start one by saying what you want in plain words – more repeat purchases, a product launch, a quiet segment won back – and a short wizard turns it into a real campaign: the AI writes the plan, breaks it into checkpoints, produces the assets each checkpoint needs, and schedules them. Every active campaign lives as a card you can open – the plan, the work produced, the status of each step – so 'what is my marketing doing right now?' has a one-glance answer. Finished campaigns keep their results, so what worked is on the record for the next one.

• Describe the outcome in plain words – a four-step wizard does the rest
• The AI writes the plan, produces the assets, and schedules the work
• Each campaign is a live card: plan, produced work, and status in one place
• Completed campaigns keep their results on the record

Email Studio – describe the email, get it in your brand template

Email Studio was rebuilt around one promise: the emails it makes are yours. Open it and describe the email right on the canvas – pick a type, write a brief, and the AI writes it in your brand voice. If you've added a brand email template in Brand Settings, the result doesn't approximate it – it IS your template: your page color, your rounded content cards, your typography, your footer, recreated as fully editable sections and blocks. The new section blocks are the Klaviyo-class building unit: each card carries its own background, corner radius, and padding, so template designs survive editing instead of flattening into a generic white column. Generation upgraded to a stronger writing model, subject lines come with scored A/B variations, and the whole studio was refinished to match the rest of the product – one quiet toolbar, a staged canvas, sequences in a full-height panel.

• Describe the email on the canvas – type chips, a brief, and one button; blocks stream in ready to edit
• Brand template fidelity: your template's page, cards, spacing, and typography recreated as editable sections
• Section blocks with their own background, radius, and padding – template designs survive editing
• Stronger writing model with scored subject-line variations to A/B test
• Faithful HTML preview and export at every step – what you see is what sends

Product Studio – one product photo in, nine art-directed shots out

Drop in a single product photo and walk away with a complete shoot. Product Studio reads the photo to understand what your product actually is, then a senior-creative-director pass art-directs every shot around it – apparel goes on a model, drinkware lands on real surfaces, hard goods get the staging they deserve. Choose the Shopify pack for a complete product page – compliant hero, second angle, detail, human scale, lifestyle – or the Ads pack for thumb-stopping feed and story creatives with headline space already reserved. Your product is preserved exactly as photographed in every frame: the photo is the source of truth, and nothing overrides it. All nine shots generate in parallel, land in your library automatically with permanent URLs, and open full-screen with one click. It's in Chat too – attach a product photo and ask for a pack.

• Nine shots per pack, each art-directed for what your product actually is – not a generic template
• Shopify pack for the product page, Ads pack for paid social with reserved headline space
• Your product preserved exactly – the photo always wins, environments change around it
• Every shot auto-saved to your library on a permanent URL; failed shots retry at no extra cost
• Works from Chat and autonomous mode – attach a photo and ask for the pack

Landing pages that convert – capture leads, publish on your domain, measure everything

The Landing Page Builder grew up from making pages to making customers. Describe your offer and the AI writes a complete direct-response page – benefit-led hero, concrete social proof, objection handling, and a working lead form – in your brand voice, grounded in what your data shows. Hit Publish and the page is live on a real URL instantly; point your ads at it, or connect your own domain right from the builder: enter it, add one DNS record, and SSL plus activation happen automatically. Every submission lands in your CRM contacts and can auto-enroll into a live journey – ad to page to contact to welcome sequence, with nothing in between. Views, clicks and form submits flow into your analytics on their own, and published pages ship with compiled styles so they load fast enough to keep your ad quality scores happy.

• Lead-capture blocks built in: submissions land in Contacts and can start a journey automatically
• Publish to a live URL in one click – or connect your own domain, fully self-serve with automatic SSL
• Page views, CTA clicks and form submits tracked in your dashboard; optional GA4 and Meta Pixel per page
• AI writes direct-response copy – one goal, specific proof, objections handled – in your brand voice
• A redesigned builder: describe the offer on the canvas and watch the page stream in, section by section

A new Dashboard – your customer lifecycle command center

The Dashboard has been rebuilt from the ground up around the thing that actually runs your business: your customers and your money. It opens with The Brief – a daily executive read written from your real revenue, channels and segments, ending in the single highest-leverage move, one click from done. Below it, everything a marketer checks in the morning lives on one screen: revenue with goal pacing, your customer lifecycle with real movement between stages, predictive insights (purchase propensity, churn risk, lifetime value and the next best action), live journeys, the full channel funnel from impressions to ROAS, and uplift – the incremental revenue your journeys actually caused, measured against a held-out control you can switch on per journey. New accounts see the whole board alive on clearly-labelled sample data, with a getting-started path that checks itself off.

• The Brief: a daily read of your numbers that ends in one recommended move – and it follows your 7/30/90-day range
• Customer lifecycle with movement: who moved up, who slipped, who entered – not just a snapshot
• Predictive insights per segment: purchase propensity, churn risk, lifetime value and the next best action, ready to run
• Uplift measurement: turn on a holdout when you launch a journey and see the revenue it actually caused
• Channel funnel: impressions → clicks → conversions → spend → ROAS per channel, with a clear scale-or-fix verdict

Your warehouse fills itself – Shopify orders and ad spend sync automatically

Connect Shopify and your orders flow into your warehouse every day – revenue, customer lifecycle and segments stay live without ever exporting a CSV. Connect Google Ads or Meta and your daily spend, impressions, clicks and conversions land in the funnel the same way. Every sync uses the same contract as an import: it appears in your Uploads list, it never double-counts, and you can remove it any time. Email engagement got real too – opens and clicks on managed sending are now tracked out of the box. And so nobody has to guess how it all fits together, a short 'How data gets in' guide now lives on the Dashboard, Connections and the Data manager – imports and integrations work together, in one warehouse.

• Shopify: one click to connect, a year of orders backfilled, then synced daily
• Google Ads & Meta: daily spend and funnel metrics – ROAS and CAC without an ads CSV
• Opens and clicks tracked automatically on managed email sending
• Load a realistic demo dataset from the Data manager – and remove it just as easily

Text your customers – through your own SMS account

SMS is here. Reach customers where they actually read – a text gets opened in minutes, not days – using your own Twilio, Telnyx, Vonage, MessageBird or SMS.to account. Your number, your sender reputation, your rates. Opt-out is built in from the first send: anyone who replies STOP is suppressed automatically, so you stay compliant without having to think about it. And SMS is a first-class step in Journeys – add a text to any automation, write exactly what it should say, and send it to the person in that run or broadcast to a whole audience. A live counter shows how many characters and how many texts each message uses as you type, so there are no surprises on your bill.

• Bring your own provider – Twilio, Telnyx, Vonage, MessageBird or SMS.to – so it's your number and your rates
• STOP and HELP opt-outs are handled for you, automatically – compliant from the very first send
• Add an SMS step to any Journey: write the text and it sends, to one person or a whole audience
• A live character and message counter as you type – including when an emoji quietly doubles the cost

Connect your whole stack – and your AI actually uses it

You can now connect over 3,400 tools to The AI CMO – the apps you already run your marketing on, across acquisition, retention, and operations. And your AI doesn't just know they're connected: it can work inside them. Connect your calendar and ask it to add a meeting; connect your project tool and ask it to open a task; connect your CRM and ask for the latest leads – it figures out what each tool can do and does it. The same is true on autopilot, so your journeys and campaigns can act across your connected tools, not just talk about them. And if you'd rather keep it simple, you don't have to connect anything at all – The AI CMO is a complete marketing platform on its own.

• Over 3,400 integrations – social, ads, email, CRM, commerce, analytics, even project management like Jira and Notion
• Your AI can act inside a connected tool, not just see it – "add this to my calendar", "open a task", "pull my latest leads"
• Works the same on autopilot – your journeys and campaigns can do real things across your connected tools
• A redesigned Connections page: your key channels up front, the long tail a search away, with Email and SMS both first-class send channels
• Open or all-in-one – bring the tools you already use, or run everything inside The AI CMO

Your Google Ads Strategist just got sharper

Auditing a live Google Ads account is smoother, and the counsel is sharper. Pick an account and you'll see it by its real name, so you always know exactly what you're looking at. And every audit now runs on the most capable intelligence we offer – the kind that reads your account the way a seasoned strategist would, and points you straight to what's worth changing.

• Pick an account by its real name, so you always know exactly what you're auditing
• A smoother, more dependable connection to your live Google Ads data
• Powered by the most capable intelligence we offer, for sharper, more strategic recommendations

Your content looks the way it should in Chat

Everything your AI CMO writes in Chat now reads clean and complete – proper headings, bullet lists, and emphasis instead of stray symbols. Social posts show inside their real platform card (LinkedIn, Instagram, Facebook, X, TikTok), and when you ask for a tweak – "make it shorter", "drop the hashtags" – the updated post stays in its preview instead of dropping to plain text. And when your AI CMO asks you a quick question back, like which image size you'd like, that now reads as a normal message rather than a half-finished post.

• Generated copy is formatted properly – headings, lists, and emphasis all show correctly, with no stray symbols
• Social posts preview in the right platform's card, and keep that preview when you refine them
• A follow-up question from your AI CMO stays a normal chat reply – no more mistaking it for a post

Your Library is easier to get around

Browsing everything you've created just got calmer and faster. Your Library now turns page by page – jump to any page, see exactly how much you have ("Showing 1–30 of 142"), and reach the bottom without an endless scroll that keeps loading. Flip pages with your left and right arrow keys, close a preview with the Escape key, and it remembers whether you prefer grid or list.

• Page-by-page navigation with a clear running count – no more infinite scroll
• Keyboard shortcuts: left and right arrows to change pages, Escape to close a preview
• Remembers your grid-or-list choice between visits
• Tidier item actions – Download only appears on things you can download (images, video, files); text and ad copy show Copy instead

Flash Report always shows the brand you're viewing

If you run more than one brand, your Flash Report now always reflects the one you have selected – its numbers, its industry, its results. A brand-new profile correctly shows an empty report until you connect its data, instead of borrowing another brand's figures. Switch brands and the report follows you.

• Numbers, industry, and results stay tied to the brand you're viewing
• A fresh brand shows a clean, empty report until its data is connected – no borrowed numbers

Every image now runs on Google's latest image model

We moved every image you generate – across the Image Center, Banner Studio, the ad creators, and anywhere your AI CMO makes visuals – onto Google's newest production image model. Nothing changes in how you work; the engine underneath is simply the latest, most stable version.

• Applies everywhere images are made – Image Center, banners, ads, social visuals, and video opening frames
• Same studios and the same steps – just a more current engine behind them

Upload one CSV, get a marketing brain – Analytics for every plan

Your AI CMO now has a memory for your business numbers, and feeding it takes one file. Drop a single CSV with your orders and your ad spend together (or separate exports from your store and ad platforms – we detect which is which) and the whole platform lights up: revenue and ROAS on the new Analytics home, customer lifetime value and repeat rates in Customer Intelligence, per-campaign profitability everywhere, and an AI chat that can finally answer "how are sales doing?" with your real numbers. Works on every plan – no warehouse required.

• One file is enough – purchases, ad spend, and funnel events (page views, carts) in a single CSV; a sample file shows the exact format and demos everything, including a money-burning campaign so you can watch the system catch it
• Three columns get you started: a customer, a date, an amount. Everything else is optional, and every extra column you bring is kept as an attribute – never lost
• Messy platform names get standardized for you – GADS, PPC, and Google Ads all become one channel; "Spring Launch 2026" and "spring_launch" are recognized as the same campaign
• Customers are created automatically from purchases, with computed properties you never upload: lifetime value, total orders, first/last purchase, acquisition source
• The Data manager (Bloomreach-style) shows every property with type, live example, and your own descriptions – plus an upload history where any import can be deleted with one click
• Re-uploading is always safe: the same order or the same spend day replaces itself, never double-counts
• Optional AI creative analysis on ad exports – your ad headlines get rated great/okay/poor into What's Working (15 credits per run)
• Acquisition, Flash Report, and the executive views all read the same numbers – channel ROAS, CAC, funnels, weekly traffic lights

Your ratings meet your revenue – campaigns finally connect end to end

Until now, what you rated and what you earned lived in different worlds. Now every campaign is one thread: content created in the platform carries its campaign identity into every tracked link and UTM, your store data and ad exports resolve back to the same campaign automatically, and Marketing Pulse shows the verdict – your team's ratings sitting right next to actual revenue, spend, and ROAS per campaign. Red means a campaign is burning money; the daily winner detection now flags both the 2x-and-up campaigns worth scaling and the silent spend leaks worth killing.

• Campaign Scoreboard on Marketing Pulse – rated count and Great % next to orders, revenue, ad spend, and ROAS, one row per campaign, red below 1x
• Copy a campaign's canonical name from the scoreboard and use it on your ad platforms – every future export then finds its way home without any matching work
• Winner detection watches the money now: campaigns returning 2x+ on spend surface as opportunities; spend with zero attributed revenue surfaces as a leak alert
• What's Working refocused on the dialogue: you rate and rank, the system answers back with detected patterns and winning playbooks built from YOUR feedback
• Unmatched platform campaigns get a one-click confirm queue in the Data tab – link them once, the answer is remembered

Campaigns – a new home for the work your AI CMO is shipping

Marketing isn't a tool catalog. So we built Campaigns – the place to start when you have an outcome in mind, not a tool. Tell your AI CMO what you want ("Acquire customers this quarter", "Build a retention program", "Launch our new product"), pick how much you want it to drive (you in the seat with chat coaching, or it on autopilot), and the plan comes together. Each campaign tracks itself through to done and lives on its own card – active campaigns at the top, completed below. Click any campaign to keep working on it; hover any card to delete one you don't need.

• Start by outcome, not by tool – 19 starters spanning the marketing funnel: Acquire customers, Convert leads, Activate signups, Retain customers, Drive expansion, Win back lapsed, Build referrals, Launch product, Build awareness, Thought leadership, Refresh positioning, Open new market, Sales sprint, Newsletter program, SEO engine, Retention program, Run an event, Build community, Customer research – or type anything in plain words
• Coach me mode (every paid plan) – your AI CMO walks you through each step in chat, drafts the work, you stay in control of what ships
• Run it for me mode (Manager and up) – your AI CMO plans, drafts, and ships in the background; you approve the key calls
• Smart channel picks – your AI CMO reads your outcome and your business profile, then suggests the right mix (paid search, paid social, SEO, email lifecycle, SMS, push, direct mail, affiliates, sponsorships, community) instead of a one-size-fits-all list. Skip if you'd rather it figure it out
• Every campaign card shows status at a glance: Running, Needs input, Paused, Planning, or Failed – plus how many steps are done and when there was last activity
• Home now surfaces your active campaigns in 'What I know about you' so you can jump back into work without hunting for it

Chrome Extension – redesigned, multi-chat with history, and you can finally stop a task

The extension that lets your AI CMO drive your real, logged-in browser got a complete polish pass. New design that matches the web app. Your conversation now survives closing the side panel – open it later and pick up exactly where you left off, mid-stream and all. Start a new chat without losing the old one (every conversation lives in a history dropdown, auto-titled by what you asked). Hit Stop to actually stop, mid-thought, mid-tool, mid-anything. And a small emerald dot on the toolbar icon tells you when something is happening in the background, even with the side panel closed.

• Full visual redesign matching the dashboard – light surfaces, emerald accents, real brand icon (not the old lightning bolt)
• Multi-chat with history – every conversation saved, switch back any time, auto-titled by the first thing you asked. The chat-bar at the top shows your current chat with a one-click dropdown for the rest
• Real cancellation – the Stop button now actually aborts the LLM call mid-stream. No more burning tokens after you change your mind
• Persistence – close the side panel mid-task, reopen later (or refresh, or restart Chrome), the conversation is right where you left it. Even works for tasks that finished in the background while the panel was closed
• Live indicator – emerald dot on the toolbar icon when work is happening, so you know the extension is alive even when you can't see it
• Elapsed time on "Working…" ("Working 12s", "Working 1m 23s") so you can tell still-going from stuck. Plus a Retry button on any failed message
• Token cache + reliability fixes under the hood – fewer auth flake errors, faster turn-arounds for back-to-back tasks

Real X threads, smarter content drafts, and bulk-content actually goes to your library

Three related improvements to how your AI CMO writes and ships content. X threads are now a first-class output – your AI CMO can plan a multi-tweet thread and publish it as a properly chained reply, not posted as one wall of text. Every content tool (social posts, blogs, emails, ad copy, landing pages) now taps into a 58-skill domain expertise library (Twitter-marketing, X-virality, brand-guardian, voice-extraction, growth-hacking, and more) so the drafts land sharper out of the gate. And bulk content – where you generate 10 posts in one shot – goes to your library for review instead of trying to publish the whole batch as a single post.

• Threads on X publish as chained replies the way X expects, not as one long blob – your AI CMO knows when an outcome calls for a thread vs a single post
• 58 marketing skills baked into every content tool – your AI CMO now writes with the same playbook strong human marketers use (channel-specific voice, virality patterns, brand consistency, founder-led storytelling)
• Bulk content lands in your library for review – pick the pieces you want to ship one by one, instead of one monster auto-published draft
• Smarter regime checks – autonomous mode now publishes via the right path based on the platform: X / Instagram / LinkedIn / TikTok / Facebook / YouTube go through your Chrome Extension (no more 40-credit cloud-browser attempts on platforms that block automated browsers), everything else stays on cloud

Your AI marketing team grew to 14 – and anyone can now preview the workspace before upgrading

Autonomous Mode got a bigger team, sharper briefs, and an open door. Three new agents fill gaps you actually have: Rebuild My CRM Strategy reads every campaign you've ever sent and writes a new approach. Win Back Customers segments your dormant CRM contacts and runs the re-engagement. Convert More Visitors audits any live page you paste in and ranks the friction by impact. Every agent description is rewritten plain-English with concrete deliverables and a 'Reads from' line so you know exactly which of your data each one uses. And anyone, on any plan, can now click 'Preview the workspace' and explore Today, Team, Office, the templates, and the hire flow before they commit.

• Three new agents: Rebuild My CRM Strategy, Win Back Customers, Convert More Visitors – 14 pre-built in total
• Every agent card shows what it Reads from (your CRM, media library, connected platforms, saved logins) and what it Needs to do good work
• Bracket placeholders like [paste your URL] or [your product name] make it obvious where to fill in for templates that need user-specific input
• Target finish date renamed and clarified – a 'X days from now' badge appears next to the date so a 90-day plan with a 7-day deadline is visible before you hire
• Agents now know today's date – they pace their work to fit your timeline, and flag the conflict in their plan if your deadline can't fit the scope
• Preview the workspace: anyone, on any plan, can now explore Today, Team, Office, all 14 templates, and the hire flow before committing to upgrade

Your AI CMO can now operate any website, even without the Chrome extension

When you ask your AI CMO to do something on a website – check Gmail, post to LinkedIn, find a contact in HubSpot, look up a competitor's pricing – it now has two paths. With the Chrome extension installed it drives your own browser using your existing logins (recommended). Without it, a built-in cloud browser opens right inside the chat, you sign in when asked, and it takes it from there. 40 credits each time the cloud browser runs, refunded automatically if it doesn't complete.

• Live preview shows the cloud browser inline as it works – watch it, or step in to log in
• Saved logins are user-scoped: tick 'save this login' after signing in and the next session for that site picks up where you left off
• Indicator in the input footer tells you which path is active – 'Extension connected' or 'Built-in browser connected', hover for details
• When a login wall appears the live view auto-opens with clear instructions
• Refund is automatic on failure (anti-bot wall, timeout, agent error) – never charged for an incomplete run

A first-run walkthrough that adapts to where you are

Your first time on /home, your AI CMO walks you through what each part of the platform actually does – the morning brief, today's first move, Chat or Autonomous depending on your plan, Studios, Library, Projects, and Track. It branches by your state: brand-new accounts get a 'finish setup' prompt, onboarded-but-unpaid get a 'pick a plan' nudge, and ready users get the full tour. Steps disappear once you're past them.

• 8 contextual steps that highlight the right part of the sidebar as you read
• Tier-aware: Intern/Starter see the Chat pitch, Growth+ see Autonomous Mode
• State-aware: each step only fires if it actually applies to you
• Post-auth now lands on /home (the brief), not a blank dashboard

/home – your AI CMO's morning brief, every day

We rebuilt the front door. /home now opens with a written brief from your AI CMO – what's working, what's next, the one move worth doing today. Below it: in-progress work, this week's scorecard, supporting cards from your real signals, and a 'try this next' shelf. No dashboards to decode.

• Editorial brief composed from your live signals – campaigns, alerts, learned patterns
• Single, concrete primary CTA tied to your current goal
• 'While you were away' overnight summary for returning visitors
• Profile switcher inline at the top so you can swap brands without leaving the page

See your AI CMO's first brief before you sign up

The landing page now ships a live demo: paste your website, get an AI CMO brief written from your actual site in the same voice you'd see every morning in the dashboard. Personalized to your business name, industry, and category – not a generic pitch. Onboarding then pre-fills from that brief so you don't re-answer what we already know.

• Reads your homepage live and writes the brief from what it finds
• Same editorial voice as /home – names your brand, calls out positioning
• Onboarding pre-fills business name + industry + URL from the brief
• Brief recap surfaces on /choose-plan so the pricing decision feels continuous

/settings is your single hub for everything account

Account, Integrations, Team, Storage, Memory, API, Billing, and Preferences – all under one tabbed hub at /settings, in the macOS System Settings pattern. Used to be scattered across /profile, /billing, /connections, /organization/settings. Brand still lives at /brand (it's a different mental model – your brand's voice, not your account).

• Eight tabs replacing four separate top-level pages
• Billing, Integrations, and Team embed as tabs without losing their standalone routes
• Avatar menu in the sidebar simplified to two entries: Settings · Brand
• Old URLs (/profile, /billing, /connections) all still work as aliases

/brand is your unified brand hub

Business profile, guardrails, visual identity, email templates, image style, and industry compliance – all under one tabbed hub at /brand. Used to live across /profile/business, /settings/brand, and /industry-hub. The profile tab also got the URL scrape + AI extract that the onboarding flow uses, so you can fill 9 fields by pasting your homepage.

• Six tabs: Profile · Guardrails · Visual Identity · Email Templates · Image Style · Industry
• Paste your URL on the Profile tab, get all 9 business fields auto-filled
• Old /profile/business and /industry-hub redirect here

Studios, reorganized by what you want to ship

Picking from 100+ tools as a non-technical marketer was the wrong starting point – marketers think 'I need to write something' not 'I need the blog-article-writer tool.' Studios now opens with six intent pills (Write · Design · Film · Plan · Launch ads · Build a page) that filter the catalog to just the tools that match. Category subheads switched from feature lists to outcome statements: 'Brand-perfect visuals in seconds, no designer needed.'

• Six intent filter pills filter the 100+ tools by what you're shipping
• Recently-used tools surface in a horizontal scroll at the top
• Outcome-focused category copy ('Run ads that convert on Google, Meta, and Amazon')
• Card layout: icon-left horizontal rows, iOS Settings pattern

/track opens with an editorial brief

Track (formerly Marketing Pulse) used to drop you into a scorecard. It now opens with an editorial brief at the top – what's working, what to watch, where to spend the next dollar – synthesized from your live data. The scorecard, alert intelligence strips, and channel detail still live below, but the answer comes first.

• Brief at the top: best/weakest channel, shipped-this-week, active strategies
• 'What's working' and 'Watch out' patterns lifted from learned_patterns
• Active strategies row with linked-content counts
• One coherent dashboard, no longer two stacked pages

A sharper engine across image, video, and ads tools

Image prompt enhancement, the UGC workflow, the demo brief on the landing, Google Ads generation, the landing-page builder, and the chrome-extension agent all moved to our newest reasoning engine. Sharper analytical pass at a lower cost per call – which means we use it more aggressively where it helps.

• 9 server-side generation paths upgraded
• Pricing: $5 in / $25 out per 1M tokens (was $15/$75 on the 4.7/4.5 generation)
• Cost logger updated so per-tool spend is tracked correctly

Strategy Creator and workflow tools think harder

Strategy Creator – plus the workflow nodes for blog articles, social posts, email campaigns, PPC ads, PPC scaffolds, and landing pages – now run on a newer, stronger reasoning engine. Better-structured strategies and stronger long-form output across the board.

• Strategy Creator upgraded to the newest generation engine
• Workflow text nodes (blog, social, email, PPC, landing page) upgraded with it
• High-effort reasoning kept on for Strategy Creator

Chat now uses what it scrapes, instead of flattening it

When you paste a URL, the chat shows you what it actually found – a homepage, a changelog, a blog post, a pricing page – and the drafts that follow are grounded in that content. A 'what's new' page becomes a post that cites your real updates, not a generic brand summary.

• Adaptive page detection – changelog, article, pricing, product, docs – picked automatically
• Detection receipt – see what was picked up and a sample of entries before the draft runs
• Social previews now match the platform you asked for
• Refinements keep the same preview template – tweak a LinkedIn post and it stays LinkedIn

Image Center got a major upgrade

Docked rails keep every control on screen while you work, prompts produce sharper agency-level results, and we've expanded the studios so you can jump from a product shot to a full campaign without leaving the page.

• Docked side rails – layers, references and prompt controls always visible
• Rewritten prompt engine tuned for commercial-grade output
• Expanded studio library – product, lifestyle, editorial, UGC

Writing Studio now illustrates your articles automatically

Draft a blog post or long-form piece and Writing Studio generates on-brief images inline, exactly where they belong in the article. No more drafting in one place and hunting for visuals in another.

• AI picks the right moments to add imagery
• Every image matches your brand style and tone
• Swap or regenerate any image in one click

Social Scheduler, rebuilt from scratch

A calendar-native planner that feels like the tool you already wished you had. Drag posts to reschedule, compose with AI, build threads, and set recurring posts – all without leaving the calendar.

• Drag-and-drop rescheduling across weeks
• AI composer with thread and carousel support
• Recurring posts for evergreen content

Landing Page Builder, rebuilt

A fully AI-native block editor. Describe the page, watch it stream in section by section, then polish and share. Every page has a public preview URL you can send to teammates or clients before going live.

• Streaming AI block generation
• Shareable preview links for feedback
• One-click export to publish

Projects, rebuilt around the way marketers actually work

Pick a tactic, get a ready-made plan, and work through it in swimlanes or a distraction-free focus mode. A slide-over panel keeps task details one click away without losing your place.

• Tactic picker with templated plans
• Swimlane board and focus mode views
• Slide-over task panel

Ask the AI to explain any tactic in detail

In your library and strategy views, every tactic now has an 'Explain this tactic in detail' button. One click and the AI walks you through the why, the how, and what to watch for – useful for client reporting or onboarding a new teammate.

Keyword Research is reliable again

Keyword Research was occasionally failing on production. We now call DataForSEO directly instead of routing through a flaky hop, so results come back every time.

Google Ads Strategist is now a conversation

Instead of a form, it's a chat. Ask questions, brief a campaign, and the strategist builds keywords, ad groups and copy for you – calling the right tools behind the scenes so the output is ready to ship.

Chat now auto-fills from your profile before it writes

Before the AI publishes, posts or sends, it now validates the brief against your business profile and fills in missing details itself. Fewer clarifying questions, fewer surprises, safer write actions.

A calmer, more premium look across the product

Landing pages, auth screens and navigation have been redesigned with a warm cream aesthetic, self-hosted Inter font and a new favicon. Pages load faster on mobile and the whole experience feels less busy.

The AI's memory got smarter

Memory now scores what's worth keeping, learns from your corrections, decays stale facts, and consolidates related notes. Result: the AI remembers what matters about your brand and stops repeating things you've already told it.

Chat responses are noticeably faster

We parallelised the slow parts of each chat turn and made reasoning conditional so simple asks come back almost instantly. Long tool chains still take time, but the time to first useful response is down significantly.

Arabic language support with right-to-left layouts

You can now work in Arabic end-to-end. The AI writes in Arabic, tools receive language instructions, and the interface flips to right-to-left where it should.

Brand Settings, redesigned

Guardrails are easier to edit, custom fonts are first-class, and the whole page now matches the rest of the product. Your brand rules are where the AI looks first before writing anything.

Creative Performance Analyzer

Drop in any ad creative and the AI reads the image with AI vision, pulls out what's working, what's distracting, and gives you a concrete rewrite brief. Useful before you spend, and even more useful after.

Marketing Pulse, redesigned with an AI executive summary

A new intelligence strip, warehouse highlights, recommended campaigns and – at the top – an AI-written executive summary of what's actually happening in your marketing this week.

Seasonal alerts now fire across the full window

Marketing Pulse seasonal alerts were only triggering on the exact day of an event. They now trigger across the whole window, so you get the nudge early enough to actually act on it.

Browser actions no longer paste truncated HTML

When the chat drove the browser to paste long HTML (like a full article), content was being cut off. Fixed – the full content is substituted from chat history before pasting.

Facebook metrics now sync for posts published from chat

Posts you published from chat weren't showing up in Facebook metrics sync. They do now, with fresh page tokens fetched on every sync.

LinkedIn publishing is reliable again

Migrated LinkedIn publishing to the new Community Management v202601 API and fixed the profile lookup that was intermittently failing. Posting to LinkedIn just works now.

Email previews in the library finally render correctly

Email HTML was rendering as plain text with trailing junk in the library detail view. Fixed by isolating the email in its own iframe – what you see is what gets sent.

'Dream for me' – let the AI propose the goal

Not sure what to work on? Ask the chat to dream up a goal for you. It looks at your business, your recent activity and what's working, then proposes a goal worth pursuing this week.

Autonomous Mode is here

Set a goal, approve a plan, and the AI runs it. Notifications keep you in the loop, access controls keep things safe, and the Chrome Extension can detect when you need a human in the browser.

Better image generation across every tool

We moved image generation to a new state-of-the-art image engine, added a structured brand style guide the AI consults before every image, and added an automatic fallback so things keep flowing when providers get busy.

A cleaner, more useful Dashboard

A new Marketing Intelligence widget, calendar hover cards, and the obsolete widgets are gone. The dashboard now leads with what to do next rather than a wall of numbers.

Batch image generation

Generate several on-brand variations in one go instead of running the prompt repeatedly. Good for A/B ideas, carousels and storyboards.

Teams, invites, SSO and a redesigned billing page

A full organisation system: invite teammates, manage roles, turn on SSO, and see a billing page that finally looks like it belongs in an enterprise tool.

Task scheduler and outcome-first welcome

Ask the AI to do something on a schedule – weekly roundups, Monday-morning briefings, monthly reports. The new welcome screen also leads with outcomes instead of tool names, so you pick what you want to achieve.

Publish to WordPress straight from chat

Write a post, say 'publish it', and it lands in WordPress – formatted, with the right metadata. You can unpublish from chat too if something needs to come down quickly.

TikTok and X (Twitter) are now connected

Two new Nango-powered integrations: connect TikTok accounts and X (using the latest v2 API) to post from chat, from Social Scheduler or from Autonomous Mode.

Flash Report now runs on real warehouse data

KPI plans, industry benchmarks, month-over-month comparisons, and exec vs. channel views sourced from the same ClickHouse aggregates – so the numbers match across every page.

Data Warehouse and Customer Intelligence, premium redesign

Enterprise-grade polish across Data Warehouse and Customer Intelligence. New query tools, funnel events, campaign insights, and a cleaner churn-risk view.

What's Working, redesigned

Rate what worked, keep the signal, and have it flow back into chat suggestions. Soft-deleted items are hidden by default and ratings now sync across pages.

Video Ads: AI Ad Creator, Mass Ads Studio and Actor Library

A complete video-ads pipeline. Create actors, compose scenes, generate a dozen variations at once, and retry any clip without losing your setup. Upgraded to the newest video engine with better progress tracking.

Chrome Extension ↔ Chat bridge

The chat can now drive your browser through the Chrome Extension. Animated progress, a stop button, and graceful fallback when you don't have the extension installed.

A premium 6-phase onboarding

New users now go through a guided onboarding that collects brand identity up front – voice, colours, reference images – so the AI can produce on-brand work from the first chat message.

Brand reference images for consistent visuals

Upload reference images and the AI analyses them to keep every generated visual on-brand – same palette, same mood, same style. No more one-off images that don't match.