Space Logo

Every change we ship to Brand OS, the operating system we use to run brands, the result we expect from it, and what happened once the numbers came in.

Goals

Client brands with a graded account change

0 → 0Early read · n=0target 2 by Dec 8

2 checks · 0 passed
  • Few unexplained contradictions in each brand's first data checkNot yet by Nov 13
  • Each client brand has an approved, carried-out and graded changeNot yet by Dec 8

No bets yet

Stop rules
  • No change proposed for either brand by Nov 24: fix the data check, not the pitch

Approval cards a week, busiest operator brand

63 (n=63) → 55Early read · n=55target 10 by Nov 30

2 bets

140 changes26 bets graded11 needles moved15 measuring

Since Sep 2026. 14 of 26 graded gave a clear answer. Couldn't tell: 7 too little data, 4 not measured, 1 other changes.

Speed and reliability improvements

  • Speed & reliabilitySpeed and reliability improvements.

A Day-10 audit for marketplace brands ranks every fix by money a month

  1. Analytics

    A Day-10 audit for marketplace brands, ranked by money a month

    A new setup for brands that sell on their own store and on marketplaces. Upload channel reports to get stock, returns, ads and per-product margin views, and an audit that prices every finding a month, with its range and inputs.

  2. AI teamMeasuring

    Desks run only on queued work, with a supply check before every model call

    Behind a per-brand switch, a desk runs only when it has work: a plan deliverable, a directive, a card you resolved, a watcher that tripped, or one daily look-around. If it lacks an input it raises one to-do and parks the work on it.

    A brand runs on 10 or fewer agent asks a week

    Expected result
    We expect approval cards a week on the busiest operator brand to fall from about 60 to 30 or fewer within four weeks of the switch going onbecause a desk that only acts on queued, question-scoped work cannot invent proposals on an hourly tour of its tools.
    How we'll know
    New approval cards a week on the busiest operator brand; success is 30 or fewer by Nov 23 with model spend per output on the lifecycle and marketing desks at least halved, disproved if cards stay above 45 a week or desks produce fewer outputs for the same spend.
    What happened
    New approval cards a week, busiest operator brand: 61 → 55 (-6) (target 30)Early read · n=55
  3. Creative

    Product posts show the real product, in a planned scene

    Every post and story is planned as a photograph first: place, moment, who is in frame, where the product sits. Each picture is checked against the product's own photos; a mismatch can't ship, and a missing photo holds the post.

  4. Creative

    A product brand's feed looks planned, not repeated

    Feed pictures carry no words, and neighbouring posts never share a shot or a place: product hero, close-up, friends together, packing from above, mood. The same person no longer turns up in every post.

  5. Creative

    Stories ask a real customer question over the brand's own photo

    Stories stop restating the brand's pitch and stop shipping as grey cards. Each asks one question a customer would ask, set in the brand's fonts on an approved photo, in copy that sounds like a person.

  6. Content & calendar

    Regenerate keeps the idea and redraws only the picture

    "Regenerate" keeps the message and redoes the picture; "New idea" starts over. A post that is waiting says why ("Waiting for today's picture budget", "Needs a product with photos") instead of "Didn't pass QA".

  7. Content & calendar

    Turning autopilot off and on always starts a fresh calendar

    A reset lifts every pause the agents had placed, clears old rejected posts from the grid and shows the new days being made right away, so a brand can no longer restart into a silently paused calendar.

  8. AI team

    Marketing tests record every variant, what shipped and the decision

    "What we learned" becomes Tests: up to four variants, the one that shipped, the result in plain words and what was decided. Each desk section has an owning agent, which can ask Engineering to change it from the section header.

  9. AI team

    Engineering agent changes reach review with every check passing

    An engineering agent's pull request keeps itself up to date, fixes its own failing checks, and asks for review only once everything passes. The Engineering desk shows where each one stands, and the ones that need you.

  10. AI team

    Engineering agents file real bugs once, with proof

    Only the live product files engineering work. A platform bug needs the real error and where it happened, the same bug from several brands becomes one task, and a fix must come with a test that fails without it.

  11. AI team

    The Insights panel shows one true row per problem

    Repeated claims fold into one row, a claim an agent stops repeating expires after a day, and a "broken tracking" claim is dropped while the store's own health check passes. A stuck desk raises one to-do naming what it waits on.

  12. AI team

    Correct a wrong agent belief from chat, and it stays corrected

    From your AI assistant you can see what the agents believe and mark a belief false once you have checked the data. It leaves every agent prompt on the next run, cannot be relearnt, and the founder's note is rewritten.

  13. AI team

    Decision grades read the last 28 days and call out too little traffic

    Decisions are graded over a 28-day window, one made with too little traffic grades "Too little traffic" instead of missed, and a change on a collection being measured waits for the window to end. Desks stop logging empty investigations.

  14. AI team

    A change that touches a protected file is handed whole to a human

    When any part of an engineering agent's change touches a protected file, the complete patch goes to a human engineer instead of a partial pull request described as the whole fix, and the task cannot be retried into a second run.

Agencies run client brands from one queue, and operator mode quiets the agent queue

  1. CreativeMeasuring

    Autopilot starts over cleanly, and product posts look like real brand posts

    Turning social autopilot off and on, or pressing Start the calendar over, now really starts over: unpublished posts close, queued work is cancelled and a fresh two-week plan is made at once. Product brands get posts that look like what consumer brands actually post: the product in a real moment, no spec sheet on the picture, a different frame each time, and stories that rotate kinds instead of repeating the feed. A post that can't be made says so with a Retry instead of an empty tile.

    Expected result
    We expect owners of product brands to approve autopilot posts as made instead of skipping thembecause the posts will look like what consumer brands actually post and a reset gives a clean calendar to judge instead of leftovers.
    How we'll know
    Share of a product brand's autopilot posts resolved by the owner that are approved as made, after the first reset; success is 40% or more, disproved if under 15%.
    What happened
    Autopilot posts approved as made: baseline 0% (target 40%)No new posts were made in the first window: autopilot was paused the moment the calendar started over, so the 4-week read restarts on Oct 26, when posting resumed. A blind side-by-side review of old and new posts is added to the same bet. Posts made after Oct 26 still showed empty studio scenes and, in one case, a suitcase that was not the real product, so a follow-up adds scene planning, a check of the product against its real photos, and stories made the same way as posts; the 4-week read restarts again when that ships, and stories are read alongside posts.
  2. Sales & leadsMeasuring

    Ads get a proper 9:16 version for Stories and Reels

    Every ad still made by Brand OS now comes with a 9:16 version with the headline inside the Stories safe zone. Each ad shows that version in Stories and Reels and the feed picture everywhere else, instead of letting the ad platform crop the words off. The launch approval card shows a Stories & Reels strip so you approve what each placement will show, and results are now stored by placement.

    Expected result
    We expect Stories and Reels placements of ads made by Brand OS to hold a click-through rate close to the same ad's feed placementsbecause the headline, product and wordmark are no longer cropped out and the platform stops re-framing the picture.
    How we'll know
    Stories and Reels click-through rate as a share of the same ad's feed click-through rate, over 10,000 or more story impressions; success is 0.7 or more, disproved if under 0.5 or if Stories and Reels get under a tenth of impressions. If no campaign made by Brand OS spends within four weeks, the verdict is couldn't tell.
    What happened
    Stories and Reels click-through as a share of feed: baseline 0.56 (target 0.70)Still measuring. The result goes here once the numbers are in.
  3. OnboardingMeasuring

    Agencies run their client brands from one queue

    An agency now runs every client brand from one queue at the top of the app, and each client approves posts from an email sent in the agency's name, with no account of their own. Every brand shows what it cost this month, in the account's own currency.

    Expected result
    We expect agencies to bring client businesses onto unfold and keep approving posts as madebecause one operator can run many brands from a single queue, clients approve from an email without an account, the product no longer asks service businesses for a store or a product, and the cost of each brand is visible.
    How we'll know
    Client businesses outside our own team with at least one post approved as made in a week; success is 10 or more by Nov 30 with at least 60% of their reviews approved as made and a serving cost per brand within the bar we set, disproved if fewer than 5 brands, under 35% approved as made, or a serving cost well over it.
    What happened
    Client businesses with a post approved as made this week: 0 → 0 (0) (target 10)Early read · n=13
  4. Brand OSMeasuring

    Run a client's marketplace business from one ledger

    An operator can now run a client brand from Brand OS: reports from Amazon, Flipkart, Myntra and Blinkit upload into one ledger, a nightly reconciliation shows the five numbers per channel, margin per product per channel and the contradictions split into explained and unexplained, and a catalogue shows every product across channels with duplicates, price differences, stock and ratings. A Clients view lets the team approve changes with a reason, send the client a weekly report built only from the ledger, and time their work. A client brand never messages the client's customers, changes prices or raises ad spend past a weekly cap.

    Expected result
    We expect each client brand's first reconciliation to leave three or fewer unexplained contradictions, and each brand to get at least one account change proposed by us, approved by its founder, carried out and graded within six weeksbecause the ledger, catalogue and report are built from rows that exist rather than from prompts, and people only read, decide and carry out.
    How we'll know
    Unexplained contradictions per client brand after the first reconciliation; success is three or fewer per brand by Nov 13. Then graded account changes per brand: one or more by Dec 8 and three or more by Jan 19, disproved if none for either brand by Dec 8.
    What happened
    Graded account changes per client brand: 0 → 0 (0) (target 1)Early read · n=0
  5. Brand OSMeasuring

    Brand OS can run a software business, not only a store

    A brand can now be marked as software, and its agents think in sign-ups, activated accounts, recurring revenue and churn instead of orders and margin. The funnel shows landing, sign-up, Instagram connected, first post live and paid for each market, Business shows the cost per activated account, and a monthly growth budget is approved once. Paid spend cannot rise until an hourly check proves sign-up works end to end and sign-up events reach the ad platform, and never past the approved budget.

    Expected result
    We expect our own software product, run as a brand, to gain accounts that activate at a cost we can see and improve week to weekbecause the sign-up event costs nothing, every funnel step is already logged, and spend is fenced by a launch check and an approved budget.
    How we'll know
    Cost per activated account: paid spend divided by new accounts whose first post went live within 7 days; success is 30 or more activated accounts in four weeks at or under ₹1,000 each, disproved if fewer than 30 on a full month's budget or over ₹2,500 each in weeks three and four.
    What happened
    Still measuring. The result goes here once the numbers are in.
  6. Social publishingMeasuring

    Instagram Stories planned by the strategy and published without babysitting

    The monthly plan now decides how many stories a brand posts and on which days, by business type and region, and the owner can change it in the content plan or posting preferences. Each story is reviewed as a tap-through phone preview and shows its reach, replies and where people left. Stories are planned and made before Instagram is connected and go out once it is; a story that needs a sticker goes up on time with a "Finish on your phone" task, and if that is missed a version with the call to action on the image is posted instead.

    Expected result
    We expect at least four in five planned story slots to publish on time once a brand switches stories onbecause the plan sets the cadence, approval is one tap or hands-off, sticker frames hand off to the owner's phone with a fallback, and the paths that dropped or stranded a story are closed.
    How we'll know
    Share of story slots the owner didn't skip whose frames all publish within two hours of their time, over the four weeks after stories are switched on; success is 80% or more, disproved under 50%.
    What happened
    Story slots published on time: 0% → 0% (0 pts) (target 80%)Early read · n=27
  7. AI teamMeasuring

    Agents only ask you when it counts

    Before an agent files an approval card, one check now asks whether you put that work on hold, whether the action can actually run today, and whether the brand already has five cards waiting. Cards left to expire count against the desk that filed them, repeat asks fold into one, and a Not applicable button on a to-do keeps the same ask away for 30 days.

    Expected result
    We expect at least half of the agent cards people resolve to end approved, up from about a thirdbecause cards that can't run, are on hold or repeat a topic are no longer created, and desks pay for the cards they let expire.
    How we'll know
    Share of resolved agent cards that end approved or carried out (an expired card counts as not approved), across live brands over four weeks; success is 50% or more with the busiest brand at 5 or fewer cards waiting at every check, disproved under 35% or above 10 waiting at any check.
    What happened
    Agent cards approved when resolved: 35% → 25% (-10 pts) (target 50%)Early read
  8. Brand OSMeasuring

    Operator mode: a quieter agent queue

    A brand in operator mode now gets far fewer asks: Instagram autopilot follows operator mode, calendar cards cover only two weeks, reworded repeats fold into one, and desks with nothing new skip their hourly run.

    A brand runs on 10 or fewer agent asks a week

    Expected result
    We expect the busiest operator brand to get 10 or fewer new approval cards a week, with 5 or fewer waiting at every twice-weekly checkbecause Instagram autopilot now follows operator mode, calendar cards cover only two weeks, reworded repeats fold into one, and desks with nothing new skip their hourly run.
    How we'll know
    New approval cards filed per week on the busiest operator brand, plus cards waiting at each Monday and Thursday check; success is 10 or fewer a week and 5 or fewer waiting at every check for four weeks, disproved above 25 in any week or above 5 waiting at two or more checks.
    What happened
    New approval cards a week, busiest operator brand: 63 → 55 (-8) (target 10)Early read · n=55
  9. Brand OSMeasuring

    Setup fills itself in and asks only for what it can't find

    Brand OS now fills brand configuration from what is connected: store domain, ad account and pixel, email list, analytics property, country, logo and business profile. Setup is one readiness list that asks inline for anything it can't pick, such as what one unit costs to make and ship, and you can answer there or in your AI assistant. Settings becomes an override page that says where each value came from.

    Expected result
    We expect a new brand to reach ready without anyone opening Settingsbecause every required value is either derived from a connected source or asked for inline.
    How we'll know
    Share of setup requirements resolved by autofill or an inline answer rather than on the Settings page; success is 80% or more, disproved if under 50%.
    What happened
    Setup values filled automatically or answered inline: 0% → 100% (+100 pts) (target 80%)Early read · n=14
  10. AI teamMeasuring

    Engineering cards need a verified reason

    The engineering desk can no longer put a false "already fixed" in front of you: closing an issue needs its fix merged or its store check passing on the latest run, otherwise the ask is refused and fed back to the desk. Work that has shipped stops coming back as a next step or a new request, requests for files that don't exist are refused, and cards tied to an issue, pull request or workflow clear within seconds of the event.

    Expected result
    We expect more of the engineering desk's cards to be worth approving, and fewer of thembecause cards built on a claim that can be checked and is false are refused before they reach you.
    How we'll know
    Share of resolved engineering cards that were approved and carried out, with expired cards counting against; success is 40% or more with 30 or fewer cards in 28 days, disproved if it stays at 20% or under.
    What happened
    Engineering cards carried out when resolved: 18% → 22.9% (+4.9 pts) (target 40%)Early read · n=35
  11. Creative

    Connect your store once and every post uses the real product photo

    One Connect Shopify button in Setup, or in a new account's first to-dos, pulls products, variants and photos into Brand OS, tags each photo (packshot, lifestyle, colour) and feeds the exact variant image into every post, with the featured product shown in the post drawer. The store's install link now lands straight on a one-click connect for the right brand. Product brands need a connected store before autopilot runs, and disconnecting pauses it with the reason.

  12. Social publishing

    Autopilot setup lists every missing step as a top to-do

    Product brands get one setup checklist on the calendar, from store connected to first post live. Each missing step becomes a top to-do that clears itself, and an agent's pause shows its reason with Lift the pause or Keep paused.

  13. Creative

    Product posts must show the real product

    Each product picture is checked against the store's photos for shape, colour, detail and scale, with no invented logos. If nothing passes after a repair and a retry, the best attempt still reaches review, marked Best try with what's off.

  14. Social publishingMeasuring

    Stories run on the days without a feed post

    Stories now fill the days a brand has no feed post, so followers see the brand on more days of the week instead of a post and a story landing together. Days an owner picked by hand are kept.

    Expected result
    We expect brands to show up on more days of the weekbecause stories stop landing on the same days as feed posts.
    How we'll know
    Share of planned story days that fall on a feed day, over the next 14 days for each brand; success is none for every brand whose stories fit in its free days, disproved if any brand on a plan- or default-set schedule still has one a week after release.
    What happened
    Story days that share a feed day: 69.2% → 8.9% (-60.3 pts) (target 0%)Early read · n=79
  15. Brand OS

    Steer your brand from your AI assistant

    Everything Setup can change, from goals and team mode to autopilot, how many cards agents may file and sales report uploads, can now be changed in a conversation with your AI assistant, under the same guards and with the reason logged. Updating the brand plan from a conversation changes only the parts you name.

  16. Analytics

    Marketplace sales and a demand scorecard

    Upload Amazon seller reports (orders, sales, settlements, search terms and returns) and Brand OS shows marketplace sales by product, which search terms to pause or raise, and a scorecard that grades a demand test against the bars you set. Marketplace fees and ad spend come off your margin, while store numbers stay as they were.

  17. AI team

    Review and merge store fixes from one engineering inbox

    The engineering view now answers four questions on one screen: is the store OK, what needs you (at most five cards), what the engineering agent is working on, and what shipped. Merge a proposed fix, send it back with a note, or close it from its card, and only the version you reviewed is merged. The agent no longer proposes fixes for checks that were skipped or failed only once.

  18. Creative

    Experimental: posts and stories can move

    An opt-in switch in posting preferences turns some photo posts and stories into a short silent clip made from the finished picture: a seamless loop, or an eight-second video. Clips are capped at one to four a week, always wait for the owner's approval, and "Use the picture instead" falls back to the still; approved feed clips publish as Reels.

  19. Sales & leads

    Pitch a prospect from its website

    A signed-in agency can scan a prospect's website and get a brand card, its Presence Score and three draft posts to show them, without the prospect signing up. Generation stops after the drafts.

  20. Content & calendar

    Generate more keeps filling the calendar ahead

    Generate more now makes the next week of posts after the last one already made, up to 75 days out, and says clearly when there is nothing left to make. One banner above the calendar says what autopilot needs from you, and every autopilot setting sits behind it in one window.

  21. AI team

    Post approvals come five at a time, soonest first

    Instagram approval cards come five at a time in posting order, and the post drawer has one clear button with quieter secondary links. The ideal customer can be written without an approved brand thesis, and new non-post approval cards no longer hide behind planned posts on Today.

  22. AI team

    Founder goals can span months and are paced fairly each week

    A founder goal can now run for up to six months and link to a public goal. The weekly review keeps its progress up to date until the end date and paces it from the day it was set, so a goal that is on track no longer reads as failing.

  23. Analytics

    Our own tools and bots are no longer counted as shoppers

    Visits from our own store checks and audits, from crawlers and from headless browsers are kept out of every funnel number, and past visits from those sources were reclassified. Agents and the leak ledger now reason about real shoppers only.

  24. AI team

    The engineering desk stops filing work its own evidence contradicts

    A tracking fix is routed to investigation when the live store check shows the event already arrives, a task promising a number the code can't move is dropped, and a leak is only called once real traffic meets the region's benchmark.

  25. AI team

    The engineering desk tells the truth about code access, CI and stuck tasks

    CI reads "unknown" with the reason when checks can't be read, a missing code-host permission becomes one to-do per repository that clears itself once granted, and a task stuck at "coding" resolves to "needs a human" with the files.

  26. AI team

    Dismiss a wrong or duplicate insight, and it stays dismissed

    An insight you know is wrong can be closed from the dashboard or from your AI assistant, with a note. Dismissed insights stay hidden even when the founder agent rewords the same claim.

  27. AI team

    Weekly reviews run every Monday, and the engineering goal shows a live number

    The weekly founder report, goal review and memory consolidation now run on a fixed weekly clock instead of being reset by every deploy, and the engineering desk goal shows its live value and verdict instead of nothing.

Autopilot posts are set in each brand's real fonts and look, and scheduled Instagram posts now go out

  1. Content & calendarMeasuring

    Posts set in the brand's real fonts, at Instagram's exact sizes

    The AI now draws only the photo, picking the best of three against the brand's own website, and the headline and logo are set on top in the brand's real font, so words are never misspelt or cut off and the logo never drifts. Every post comes out at Instagram's exact size for feed, story or Reel. The quality check only holds a post back for problems that are clearly wrong, such as a missing logo, unreadable text or stray words; anything else reaches you for review with a note.

    Expected result
    We expect owners to approve most of the posts that reach thembecause spelling, cropping and logo mistakes can no longer happen and posts are no longer held back over matters of taste.
    How we'll know
    Share of posts that reach review and are approved rather than skipped by the owner, over 4 weeks; success is 50% or more, disproved if under 20%.
    What happened
    Reviewed posts approved by the owner: 0% → 27.7% (+27.7 pts) (target 50%)Early read · n=65
  2. Content & calendarMeasuring

    Every post is composed in the brand's real look from a real idea

    Each autopilot picture is now made in one go, headline and logo included, from that day's content idea, with the brand's website screenshots, logo, a sample of its real font and, for product brands, its product photos as references. Two versions are made and checked against the exact words, logo and product; a post that still fails is held back instead of published. Editing a headline changes the words on the same picture.

    Expected result
    We expect far more autopilot posts to reach review, and some to be approved as madebecause the AI now sees the brand's real website, font and products and is given a real idea and the exact words to use.
    How we'll know
    Share of autopilot posts that reach review instead of being held back, over 28 days; success is 70% or more with some approved as made, disproved if it stays under 40% by Nov 13.
    What happened
    Autopilot posts reaching review: 20% → 93.7% (+73.7 pts) (target 70%)Enough data · n=570
  3. Social publishingMeasuring

    Scheduled Instagram posts go out on time, or say why up front

    Posts from the content calendar and autopilot now publish at their scheduled time, or straight away with Publish now: single pictures, carousels of up to ten, and Reels. Scheduling runs the same checks Instagram does (file type, size, shape, video length, caption and hashtag limits, account health and the daily cap) and lists every problem with its fix at once, so nothing fails later. Times follow the brand's own time zone.

    Expected result
    We expect every scheduled Instagram post to either go out within minutes of its time or be stopped when it is scheduled with a reason you can act onbecause the checks now run at scheduling time and scheduled posts are now actually sent.
    How we'll know
    Share of scheduled posts published within 5 minutes of their time; success is 98% or more by Nov 5, disproved below 90% or by any post left unpublished with no reason.
    What happened
    Scheduled posts published on time: 0% → 75% (+75 pts) (target 98%)Early read · n=4
  4. OnboardingMixed

    Add a brand in three steps, or run it as a marketing agent

    Adding a brand is now a three-step wizard: choose how you'll use Space AI and scan the brand's website, confirm or upload the logo, then approve the color palette and design profile before the brand is created. Brands set up for marketing open into a simpler five-tab view (calendar, creative, paid marketing, brand and audience, setup) with the social calendar as home.

    Expected result
    We expect new brands to finish their logo and design setup during onboarding and reach their social calendar the same daybecause the website scan, logo and palette are approved in the wizard instead of later in setup.
    How we'll know
    Share of newly added brands with an approved design profile and logo when they are created; success is at least 80%, disproved if fewer than 40% finish the website scan step.
    What happened
    New brands with a ready design profile and logo when created: 0% → 50% (+50 pts) (target 80%)Early readEvery brand added since release had its website scan and design profile ready when it was created, but only half had a logo. Most new brands now arrive through the self-serve sign-up, where the logo check became a to-do after landing rather than a step in the way.Checked Oct 26, 2026
  5. Content & calendarCouldn't tell · Too little data

    Your social calendar plans 45 days of varied posts

    The calendar now plans and makes the next 45 days of Instagram posts, always rolling from today. Each post gets a type (product hero, lifestyle, feature callout, question, festival, waitlist teaser and more), one of the brand's core messages, and a headline set in the brand's own font when the post calls for one. Pictures are built from up to four of the brand's real product photos plus its own Instagram posts as style references, and the content plan behind it is yours to edit.

    Expected result
    We expect more autopilot posts to be approved as they were madebecause pictures show the real product, on-image text matches how the brand already posts, and posts vary by type and message.
    How we'll know
    Share of posts that reach review and are approved without a redo or a product swap; success is at least 60%, disproved if it is under 30% after three weeks at three or more posts a week.
    What happened
    Graded early because product posts are moving to the old app's showcase flow. By 15 October only one autopilot post had reached review and none was approved, too few to judge. Most posts stopped at the quality check, and across all creative work 1 of about 386 pictures made in four weeks was approved.
  6. Onboarding

    Autopilot is ready as soon as you add a brand

    Adding a brand now really scans its website: its colours, logo, the fonts it actually uses, a business profile and a marketing strategy, all editable in Brand settings. The wizard asks whether the business sells products, services or software, then fills in the brand story, ideal customer, content plan and strategy from the scan, so autopilot can be switched on straight away. Service and software businesses no longer need products to post.

  7. Onboarding

    Retail stores get their own business type

    Alongside products, services and software, a brand can now say it runs a retail store, in the Add brand wizard or the content plan, with an optional store address. Retail brands get everything product brands get, and their posts read like a shop's: new arrivals and in-store moments, ending with an invitation to message or visit the store, without passing off another brand's product as their own. The website scan suggests retail when a site clearly belongs to a physical store.

  8. AI teamMixed

    One approval card per real decision

    When an agent asks again for something already waiting, the request joins the existing card ("Asked 14 times") instead of adding another, and cards that have gone stale or are already done close on their own. On a card for a paused ad draft, you can let the marketing agent create similar paused drafts on its own for 7, 14 or 30 days, capped at 10 a day and withdrawn if one fails or you undo one.

    Expected result
    We expect the approval queue to shrink to a handful of real decisions and founders to start deciding againbecause repeat asks fold into one card and routine paused drafts no longer need a click.
    How we'll know
    Approval cards waiting on the busiest brand; success is 20 or fewer for 7 straight days by Oct 21, disproved if more than 40 are still waiting two weeks after release.
    What happened
    Approval cards waiting, busiest brand: 126 → 47 (-79) (target 20)Enough dataOn the brand whose queue started this, waiting cards fell from about 125 to about 20 within a week and to a handful by Oct 26, but never held at 20 or fewer for seven straight days. The busiest brand is now a new self-serve account with about 45 post approvals waiting, one per planned post for the next six weeks, so the bar was not met across brands.Checked Oct 26, 2026
  9. Content & calendar

    A monthly creative strategy and a brief behind every post

    Each month autopilot writes a creative strategy you can edit, approve or regenerate: customer insights, one campaign idea and a few named recurring series. Every post starts from a short brief; three ideas are written and scored blind, only the best is made, and a creative-director review checks it before you see it. A day with no idea good enough waits for the next planning run instead of getting a generic post, and service and software businesses get plans built from their offerings.

  10. Social publishing

    Instagram Stories get their own schedule

    Alongside feed posts, a brand can switch on a separate Stories schedule with its own days, time and number of frames. Story frames are made at full-screen size and show on the calendar next to that day's post.

  11. Social publishingDidn't work

    Failed posts tell you what to do next

    The autopilot card is a weekly posting grid: tick the days and the number of posts a week follows. Calendar posts are picture-first tiles with a clear status. When a picture fails the quality check, the post says what went wrong, recommends one action, and offers the best attempt, which goes to you for approval and is never published on its own.

    Expected result
    We expect autopilot pictures to pass the quality check within three attemptsbecause the picture instructions no longer contradict the product or the check.
    How we'll know
    Share of autopilot posts that pass the quality check within three attempts; success is at least two of the next three posts, disproved if two or more still use up every attempt.
    What happened
    Autopilot posts passing the quality check: 50% → 14.3% (-35.7 pts) (target 67%)Early readTwo of the first three autopilot posts after release used up every attempt, which disproves it. Across all autopilot posts made in the five days after release, about one in seven passed the quality check within three attempts, down from about half before.Checked Oct 17, 2026
  12. Social publishing

    Autopilot never posts twice, and a failed post can be retried

    An approved post can never go to Instagram twice, even if publishing is interrupted. A post that fails to publish says why in plain words, such as Instagram needing to be reconnected, and offers a one-click retry. Turning autopilot off stops every scheduled post, a skipped day can be brought back, and sold-out products are never featured.

  13. Social publishing

    Connect Instagram without leaving where you were

    Instagram's sign-in now opens in a new tab, so Brand OS stays on the page you started from, such as the content calendar. When you finish, the sign-in tab closes itself and Setup shows Instagram as connected within seconds, instead of saying it wasn't connected for up to six hours.

  14. Content & calendar

    Choose which products appear in your posts

    In the content plan, every product in the rotation has a Leave out button, and left-out products can be put back. Products sold in several sizes now show the size, so similar rows can be told apart.

  15. Content & calendar

    The content plan opens as one focused screen

    On the calendar, one row now shows this month's main idea and flags a draft that still needs your OK. Clicking it opens the whole content plan as one wide sheet: the month's idea leads as a large, editable headline with Approve right under it, and every section (customer, series, formats, look and products) sits below with its title on the left. Save appears only when you have unsaved edits.

Instagram on autopilot: every posting day fills with a checked post, ready to approve

  1. Social publishingDidn't work

    Instagram posts planned, made and scheduled for you

    Pick how often to post and on which days, and every posting day on the content calendar fills with a ready post: a photorealistic photo of one of your own products with your logo on it, plus a caption. Each image is checked for the logo, the exact product, nothing that isn't yours and a believable scene, and is remade up to five times until it passes. Approve a post and it goes out on Instagram at 5 PM your time.

    Expected result
    We expect brands with autopilot on to post on at least nine in ten planned posting days without making any images by handbecause every posting day gets a ready, checked post days ahead instead of waiting in a review queue.
    How we'll know
    Share of planned posting days (not skipped) that publish, over the four weeks after a brand turns autopilot on; success is at least 90%, disproved if it stays below 60%.
    What happened
    Planned posting days that publish: 0% → 11.8% (+11.8 pts) (target 90%)Early readOf the planned posting days since early October that weren't skipped, about one in ten published: two posts, both within seconds of their time. The rest were missed because nobody approved the post before its slot, mostly on new self-serve accounts that had not connected Instagram yet.Checked Oct 26, 2026
  2. Social publishing

    Open any post to see what went into it

    Each post shows its image, caption and posting time, the product and measurements it had to match, every attempt that failed the check and why, and the brand story, customer and design rules the AI worked from. From there you can edit the caption, remake it with a note, swap the product, move it, skip it or post it now.

  3. Creative

    Images use each size's real measurements

    Product details now come from each size in your store rather than only the main product, so a medium case is described and drawn at medium-case dimensions instead of cabin size.

  4. AnalyticsCouldn't tell · Other changes at the same time

    Store visit tracking records every event again

    Some events the store sent were being turned away because of how their names were written, so product-page views stopped being counted and parts of the shopper journey read as near zero. They are accepted again.

    Expected result
    We expect product-page views to be counted at their earlier level againbecause the store kept sending them but they were being rejected.
    How we'll know
    Share of store visits recorded as viewing a product page, over the 7 days after release; success is back near nine in ten, as before the drop, disproved if it stays below a quarter.
    What happened
    Visits recorded viewing a product page: 19.9% → 71.9% (+52 pts) (target 89%)Early readProduct-page views are counted again: in the first full week after the fix, about seven in ten visits viewed a product page, up from one in five before. That is short of the nine in ten we aimed for, but paid traffic was paused the day after release, so visits fell to about an eighth of their earlier level and the visitor mix changed.Checked Oct 20, 2026
  5. AI teamMixed

    Agents can tell more of their ad changes apart

    When the AI grades whether one of its ad changes worked, changes the ad platform makes on its own, such as automatically created audiences, and our own moderation of cards no longer count as 'something else changed'. Edits a person makes to budgets or campaigns still do, so a grade is only marked 'can't tell' when it really can't.

    Expected result
    We expect far fewer agent decisions graded 'can't tell'because every one graded in the week before was marked that way, mostly over the ad platform's own automatic changes.
    How we'll know
    Share of graded agent decisions marked 'can't tell' after release; success is under half by Oct 20, disproved if it stays above 90%.
    What happened
    Agent decisions graded 'can't tell': 100% → 75.5% (-24.5 pts) (target 50%)Early readAbout three in four agent decisions graded since release were still marked 'can't tell', down from all of them but above the half we aimed for by Oct 20. In the last four days it fell to about one in three, as more decisions got a clear yes or no.Checked Oct 20, 2026
  6. AI team

    Repeat requests join the card already waiting

    When a desk asks again for the same engineering fix in different words, the request now joins the card that is already waiting instead of adding another. Old alerts that say images are stuck in their quality check clear on their own once nothing is stuck.

A daily AI stand-up, shoppers who stop at checkout stay reachable, and agents learn from their own ad changes again

  1. AI teamMixed

    Your AI team holds a daily stand-up, and Slack only pings you when it matters

    Every morning at 9:15 your time, the six AI desks hold a text stand-up. Each desk says what it did, what's next and what blocks it. Any desk can pitch a change to how the brand runs, such as a new tool, more budget, an engineering fix, an experiment or a new way of working, but only by proving the bottleneck with a real number. It also has to say what the change costs, what it should move and when to stop. The other desks back or challenge each pitch with evidence, and desks whose past bets worked carry more weight. One desk a day pitches something outside its lane and another plays devil's advocate. The founder agent adopts at most two pitches, which arrive as normal approval cards. Each adopted pitch is graded when its deadline passes. You get a short, human recap in Slack, with the transcript and a 'who's been right' table on the Work tab. At 9:00 a daily founder report replaces the six-hourly brief. It covers what we did, what went well, what went badly and what we changed, what we learnt, what we want to try, and your move. Only urgent things ping you straight away, at most three a day: a guardrail breach, checkout down, a spend anomaly, or an approval worth 10% or more of monthly spend. Everything else is folded into the report.

    Expected result
    We expect founders to act on the few pings that matter much faster, and the AI team to surface fixes that get approved and workbecause Slack drops from dozens of repeated notices a day to a handful of high-signal posts, and every pitch now has to prove its bottleneck and is graded afterwards.
    How we'll know
    Share of urgent pings whose card is resolved within 24 hours, and Slack posts per brand per day; success is at least 70% resolved within 24 hours and at most 4 posts a day, disproved if resolution stays at today's rate or posts stay above 8 a day three weeks after release.
    What happened
    Half of it held. Slack sent about three and a half messages a day, under the bar of four. But only about one in seven urgent pings had its card dealt with within a day, far short of the 70% bar.Checked Oct 17, 2026
  2. AI teamDidn't work

    An approval queue you can clear, with bad proposals stopped first

    Every approval sits in one list with the most money at stake first. Contradicting proposals for the same campaign become one card with options, repeated asks fold under the original, and cards expire after 72 hours or close themselves when their cause clears. Before any card is filed, it is checked: account and campaign numbers must be your own, shopper links must be on your domain and load, and dates must be in the future, so a proposal with made-up details never reaches you.

    Expected result
    We expect the approval queue to shrink by at least two thirds and more cards to be approved as proposedbecause most cards were duplicates, routine housekeeping or contradictions, and proposals with made-up details are now refused before they are filed.
    How we'll know
    Pending approvals, and the share of decided cards approved as proposed, over the two weeks after release; success is the queue down by two thirds and approvals as proposed rising from under half to at least 70%, disproved if the queue grows back past 60% of its old size within two weeks.
    What happened
    Approvals waiting on you: 22 → 133 (+111)Early readThe queue grew instead of shrinking. Two weeks after release it was about six times its size at release, far past the 60% bar, and most of the new cards were the same few ad-setup requests filed again and again. People approved about four in ten of the cards they decided, short of the 70% goal.Checked Oct 14, 2026
  3. CustomersIndia brandsCouldn't tell · Too little data

    Shoppers who stop at checkout stay reachable by phone

    Shoppers who reach checkout but never type an email are no longer lost. The phone number they enter at checkout is now kept, the checkout is linked back to the ad visit that brought them, and they count as reachable on WhatsApp. Shoppers can also leave just a WhatsApp number on the welcome offer and waitlist, and ad platforms get more visitors they can match.

    Expected result
    We expect most shoppers who abandon checkout to become reachablebecause the phone number they already enter at checkout is now kept and tied to their visit.
    How we'll know
    Share of abandoned checkouts with an email or WhatsApp number, from about one in eight to at least 70% once 20 or more checkouts have been abandoned; disproved if it is still below 40%.
    What happened
    Abandoned checkouts we can reach: 12.5% → 16.7% (+4.2 pts) (target 70%)Early readFewer than the 20 abandoned checkouts the bet needed came in over 30 days, and none has started since paid ads paused on Oct 9. Of those there were, the share we can reach rose from about one in eight to about one in six.Checked Oct 17, 2026
  4. AI teamDidn't work

    Agents learn from their own ad changes again

    Business › Overview now opens a 'What happened this week' panel under the headline numbers. Each row is one action in plain English: who did it (an agent, you, the ad platform or a stock sync), the approval behind it, and what the ad platform did next, with a warning when your change overlaps a test an agent is running. The ad platform's routine reactions to an agent's own edit, such as reviewing an ad or starting delivery, no longer mark that agent's result as 'can't tell'.

    Expected result
    We expect the share of agent decisions graded 'can't tell' to fall sharplybecause routine ad-platform reactions to an agent's own edit no longer count as an outside change.
    How we'll know
    Share of graded agent decisions marked 'can't tell', from about five in six over the 30 days before to at most a third within two weeks of release; disproved if it stays above 60%.
    What happened
    Agent decisions graded 'can't tell': 84% → 100% (+16 pts) (target 35%)Early readTwo weeks after release, every agent decision graded in the latest window was still marked 'can't tell', against a goal of at most a third. Grades ran from about five in six before release to all of them after.Checked Oct 14, 2026
  5. Social publishingCouldn't tell · Too little data

    Connect Instagram in one click

    Setup now has a Connect Instagram button, and so does the Instagram calendar while no account is linked. One click takes you through Facebook's permission screen and back to Setup with a note saying whether it worked and, if not, why. Before, linking an account needed a manual change on our side, so the planned posts could not publish.

    Expected result
    We expect brands to publish organic Instagram posts at the planned three to five a weekbecause connecting the account, the only blocker, is now one click.
    How we'll know
    Instagram posts published through approvals per brand per week; success is at least three a week within two weeks of connecting, disproved if fewer than one a week is published once connected.
    What happened
    Instagram posts published a week: 0 → 0 (0) (target 3)Early readOne brand connected Instagram, and no autopilot post was approved in the window, so nothing was ready to publish. Scheduled posts also only started going out reliably with the Oct 14 release.Checked Oct 17, 2026
  6. CreativeDidn't work

    Upload your logos; AI decides which one goes where

    Setup now takes up to two logo uploads, typically a monogram and a wordmark, instead of a link to a hosted image, or imports your logo straight from your storefront. AI tags each one with its style, the backgrounds it works on, its script and where to use it (store header, square avatar, feed posts, stories, email), and you can edit the tags. Creative renders, quality checks, your AI team and the Brand OS header then use the mark that fits each format, such as the monogram on stories and the wordmark on feed posts.

    Expected result
    We expect brands to stop being blocked on a missing logo and creative to stop failing its logo checkbecause uploading takes about 30 seconds with no hosting, and each format gets a mark that fits it.
    How we'll know
    Open missing-logo blockers, and the share of new images that pass the logo check; success is no open logo blockers within a day of upload and at least 95% of new images passing, disproved if the pass rate stays below 85%.
    What happened
    Logo blockers cleared: both logos were uploaded and tagged within seconds, and the one setup task that was filed again after upload closed the same day. But only about a quarter of new images passed the logo check in the two weeks after upload, far below the 85% bar, and in the second week almost none did.Checked Oct 14, 2026
  7. AI teamMixed

    Engineering desk shows what's broken and whether fixes worked

    One screen now answers three questions: is your live store broken, what needs you, and did last week's fixes work. Hourly checks of your storefront file one task per failing check, fixes move through clear stages from investigating to shipped, and every shipped fix is graded against the metric it promised to move. Duplicate bug reports fold into one card, and the desk points at the real blocker instead of a stale permissions warning.

    Expected result
    We expect storefront problems to be fixed faster and fewer fixes to be shipped without proofbecause breakages are caught by hourly checks, duplicates no longer split the work, and each fix is graded on the result it promised.
    How we'll know
    Share of shipped fixes graded within 28 days, and time from a failing store check to a merged fix; success is every shipped fix graded and failing checks fixed within a week, disproved if fixes keep shipping ungraded or store checks stay red for more than a week.
    What happened
    Shipped fixes graded on the result they promised: 0% → 13.3% (+13.3 pts) (target 100%)Early readFailing store checks were fixed well within a week: none stayed red for more than about half a day, and fixes merged within one to three days. But only 2 of 15 shipped fixes were graded, both 'can't tell', and most of the rest never promised a metric, so they can never be graded.Checked Oct 20, 2026
  8. AI teamCouldn't tell · Not measured

    One list for every decision, and asks land in the right place

    On Today, every approval sits in one Needs your decision list, most money at stake first and five per page, and each section says what it is for: approve and the agent does it, or only you can do this. Agents stop turning hand work, such as uploading a logo or switching on an email flow, into engineering requests or approvals that do nothing. Those land once in Action required.

    Expected result
    We expect asks filed in the wrong place to drop to zerobecause one rule now says where each kind of ask goes and obvious misfilings are moved automatically.
    How we'll know
    Engineering requests that turn out to be hand work, and approvals that are really chores; success is both staying at zero, disproved if either keeps recurring at the old rate.
    What happened
    We don't record when an ask is moved because it was filed in the wrong place, or when an engineering request turns out to be hand work, so we can't count misfilings before or after the change.Checked Oct 14, 2026
  9. Brand OS

    An honest state for every brand, and Today opens on what matters

    Each brand now reads No sales yet, Leaking, Selling or Profitable, worked out from its orders, ad spend and margin, with a plain reason such as no orders in 28 days, instead of showing healthy just because it is live. Today opens with one line of orders, spend and margin for the week, then the funnel step where shoppers drop off most. Margin figures are marked as estimates until product costs, shipping and returns are filled in.

  10. Brand OSMixed

    Run Brand OS from your AI assistant

    Your desktop AI assistant can now read everything in Brand OS and act on it: the daily briefing, funnel, marketing, customers, the approval queue and your AI team. It uses the same context your AI founder sees. Anything that changes the business still arrives as an approval card with the usual guardrails, and each co-founder has their own key, so every approval and change is recorded under their name. Ready-made shortcuts cover the daily standup, approval triage, funnel diagnosis, the weekly review and launch planning, and money is shown in each brand’s own currency and number format.

    Expected result
    We expect founders to run their daily Brand OS work from the assistant they already usebecause it now sees the full picture and can act safely through the same approvals.
    How we'll know
    Days per week each co-founder works in Brand OS through the assistant, and the share of approvals resolved there; success is use on five of seven days with at least 30% of approvals resolved there, disproved if use stops after the first week.
    What happened
    Approvals decided from the AI assistant: 0% → 62.1% (+62.1 pts) (target 30%)Early readAbout six in ten approvals that people decided since release were decided from the assistant, double the 30% goal, and it was still in use in the second week. We don't record days when the assistant is used only to read, so we can't tell whether founders use it five days a week.Checked Oct 14, 2026
  11. AI team

    Every live ad's landing page is checked daily

    Each day, every active ad's destination is opened and checked: a broken page, a homepage instead of a product, or a link off your domain files one engineering task per ad, shown on the engineering desk.

  12. CreativeDidn't work

    The Creative desk shows what needs your review

    Work › Creative opens on the work waiting for you, says what needs you and when the next piece retires, and opens images and videos full screen so you can approve or retire them and move to the next. It also shows the Instagram plan as a month calendar with likes, comments and reach on past posts, and an Instagram section with followers, reach and top posts.

    Expected result
    We expect founders to review new creative before it goes stalebecause the desk now opens on a review queue with a retire date on each piece and a large view to review in.
    How we'll know
    Share of creative approved or retired by a person rather than retired automatically after 7 days; success is at least 70%, disproved if it stays below 30%.
    What happened
    Creative reviewed by a person, not retired automatically: 72% → 0.4% (-71.6 pts) (target 70%)Early readIn the two weeks after release, a person reviewed one new piece of creative. Every other piece, a few hundred in all, retired automatically after going stale, far below the 30% bar.Checked Oct 14, 2026
  13. Sales & leads

    The budget panel shows only what is live

    On the Marketing desk, Where the budget sits now lists only budgets that are live on Meta, with the live daily total, and leads with ad spend so far this month. Paused ad sets no longer show their parked budgets.

  14. AI team

    Platform problems go to engineering, not your to-do list

    When Brand OS itself breaks, for example images stuck waiting for their quality check, the AI desks now file a fix with engineering instead of asking you to do it, and a fix attempt that crashes is retried once on its own. Requests only you can act on, like reconnecting an account, still come to you. Images that were stuck in the quality check are being graded again.

AI desks now plan and grade their own weeks, and price changes stay inside bands you approve

  1. Brand OSCouldn't tell · Too little data

    Price bands for every product

    Set a floor and ceiling price for each product. Your AI founder can propose price changes in your store, and each one is checked against the band, your unit economics and the current live price before it is applied with your approval. Every change is listed in the Money view.

    Expected result
    We expect better margin without risky price swingsbecause the AI can adjust prices, but only inside limits you set and only after checking unit economics.
    How we'll know
    Contribution margin per order on banded products before and after price changes; success is higher margin with conversion holding steady, disproved if conversion drops enough to lower total margin.
    What happened
    No product has a floor or ceiling price set yet, and the AI has not proposed a price change, so there is no before-and-after margin to compare.Checked Oct 9, 2026
  2. AI teamCouldn't tell · Not measured

    Each AI desk owns a function and a target

    Each AI desk now has a charter: its mission, the metric it owns against a target, and its standing duties. Every Monday it plans up to three deliverables, one experiment and a prediction you can amend, and on Friday it grades its own week. Desks wake on real events such as orders, checkouts and approvals instead of every hour.

    Expected result
    We expect less busywork and more finished deliverablesbecause each desk commits to a short weekly plan against a metric it owns and grades itself on the result.
    How we'll know
    Share of agent activity that produces no deliverable, and share of weekly plans graded as delivered; success is busywork falling compared with the old hourly mode run side by side for a week, disproved if it does not.
    What happened
    In the first fully graded week, desks finished 11 of 15 planned deliverables and 3 of 5 desk predictions hit. The old hourly mode was never run alongside for comparison, and total agent decisions kept rising, so we can't say whether busywork fell.Checked Oct 9, 2026
  3. CustomersCouldn't tell · Too little data

    Customers tab becomes a full CRM

    See your ideal customer next to who is actually showing up, and work one row per person with their full history: visits, carts, orders, emails, WhatsApp messages, calls and notes. From the row you can change consent (synced to your email platform), send an approved WhatsApp template or log a call, and a STOP reply withdraws consent automatically.

    Expected result
    We expect more repeat and recovered ordersbecause every captured shopper becomes a reachable, consented contact with their history in one place instead of an anonymous visit.
    How we'll know
    Share of captured shoppers reachable with consent, and share who order after a message; success is both rising month over month, disproved if messaged contacts convert no better than those not messaged.
    What happened
    Captured shoppers reachable with consent: 19% → 54% (+35 pts)Early readThe share of captured shoppers who have given consent rose from about one in five in August to just over half in September. No messages have been sent from the customer list yet, so we can't tell whether messaged contacts order more.Checked Oct 9, 2026
  4. CreativeDidn't work

    Every creative passes a quality check first

    AI now draws only the scene; your exact headline, body text and real logo are placed on top, so nothing is misspelled. Each image is checked for the logo, approved text and claims only, and a match with the real product before it is saved, and internal product codes or persona names never appear.

    Expected result
    We expect more creatives approved without editsbecause text and logo are placed exactly and broken images are caught before you see them.
    How we'll know
    Share of generated images that pass the automatic check, and share approved without changes; the bar we set is at least 90% passing the check before a creative pack can be signed off.
    What happened
    New images passing the quality check: 90% → 26.7% (-63.3 pts) (target 90%)Early readOnly about one in four new images passed the automatic check, far below the 90% bar. Most failures were the logo: an invented mark, or the wrong logo for the format. None of the images that passed was approved by a person; they were retired after 7 days or are still waiting.Checked Oct 9, 2026
  5. AnalyticsCouldn't tell · Not measured

    One number per metric across every tab

    Test and internal orders are left out of every revenue and customer metric, and "last 7 days" means the same rolling week everywhere. Brand health is split into a Business score and a separate Data confidence score, so a brand with no orders can no longer look healthy.

    Expected result
    We expect founders and agents to trust and act on the numbersbecause the same metric no longer shows different values on different tabs.
    How we'll know
    Number of metrics that disagree between tabs for the same period; success is zero known mismatches, disproved if they reappear.
    What happened
    The automatic check that compares our order count with the store's flagged one mismatch, caused by a test order, and it was fixed the same day; none since. Other numbers are not compared across tabs automatically, so we can't confirm zero mismatches everywhere.Checked Oct 9, 2026
  6. Brand OSDidn't work

    Today opens with a daily founder's note

    Today starts with a daily note from your AI founder (what we did, what went wrong, what we learnt), each point linked to its evidence, then where the brand stands against its plan and one "Your move" list. Action items check themselves when you complete them, duplicate asks collapse into one card, and outdated to-dos close on their own.

    Expected result
    We expect founders to clear their queue fasterbecause each real ask appears once and resolved ones disappear without anyone closing them.
    How we'll know
    Cards on Today per real ask; success is about one card per ask, disproved if reworded duplicates keep stacking up.
    What happened
    Exact repeats now fold into one card, but reworded duplicates still pile up: the open queue held about 1.8 cards per real ask, such as three cards to merge the same fix and several cards about the same product page problem from different desks.Checked Oct 9, 2026
  7. AnalyticsIndia brandsCouldn't tell · Too little data

    Returns-to-origin measured from real orders

    Returns-to-origin (RTO: parcels shipped and sent back undelivered, mostly refused cash-on-delivery orders) are now measured from your shipped orders. Until enough orders settle, margin is shown before returns instead of waiting on a manual estimate, and unit costs entered once in Settings flow into margin everywhere.

    Expected result
    We expect every India brand to see margin after shipping and returns from the startbecause it no longer depends on someone typing in a return rate.
    How we'll know
    Share of India brands with unit costs that see a margin figure; success is all of them, with the measured return rate taking over once 20 shipped orders have settled.
    What happened
    The India brand has unit costs entered, but it had no shipped orders in the period, so the measured return rate had nothing to settle and we can't tell whether it would take over from the estimate.Checked Oct 9, 2026
  8. AI team

    Agents ask for the channels they're missing

    Each desk compares itself with a playbook of direct-to-consumer channels (WhatsApp for India, SMS for the US, organic Instagram, customer-style video) and can file one costed request a week. Approving it creates a setup to-do that closes itself once the connection works.

  9. AI team

    Founder instructions are checked before they go out

    Before your AI founder's instructions reach the team, they are checked for mistakes such as blaming the wrong step of the funnel, undoing a test you launched, or using claims you haven't approved. Each instruction expires after 72 hours unless renewed, and desks can challenge it.

  10. AI team

    Initiatives: multi-step plans you approve

    Your AI team can propose an Initiative, a multi-step plan you can approve, revise, pause or stop. Steps you pre-approve run within the budget caps and scope you set; everything else still comes to you.

  11. Social publishing

    An organic Instagram calendar

    • Your AI team can plan an organic Instagram calendar where every post is an approval card
    • You can edit the caption and post time right on the card

Chat with your AI founder about your brand and approve its actions in the conversation

  1. AI teamCouldn't tell · Not measured

    Chat with your AI founder

    Ask your brand's AI founder about performance or strategy in saved conversation threads. You can watch its reasoning steps as it works, approve the actions it proposes without leaving the chat, and answer when it asks you for input.

    Expected result
    We expect founders to get answers and act fasterbecause they can ask about the brand directly and approve actions in the conversation instead of hunting across tabs.
    How we'll know
    Share of rated answers marked helpful, and share of replies that finish without an error; success is most rated answers marked helpful, disproved if more are marked unhelpful than helpful.
    What happened
    Early conversations are happening, but no answer has been rated yet and we don't yet record per-reply errors, so neither part can be measured. We'll make rating an answer more visible.Checked Sep 29, 2026
  2. AI team

    Creative and support agents check live data first

    Before writing new briefs, the creative agent now reviews how your live ads are doing (creative fatigue, click-through rate, return on ad spend) and your video library, and the support agent reviews open tickets before it acts.

A weekly market briefing and a customer avatar now guide your AI team

  1. AnalyticsCouldn't tell · Too little data

    A weekly market briefing

    Each week you get a market briefing that answers the core questions: market size, how many category purchases happen each week, channel mix, seasonal demand windows, competitors in your price band and where you can win. Every number shows its sources and method, and anything the data can't back is marked unknown instead of guessed.

    Expected result
    We expect better product, price and channel decisionsbecause founders and agents plan against sourced market numbers instead of guesses.
    How we'll know
    Share of briefing numbers backed by a cited source or openly marked unknown, and how often agent plans cite the briefing; success is no unsourced number in any briefing.
    What happened
    Only one sourced market figure has been recorded so far, for a single pilot brand, and we don't track when agent plans cite the briefing. Too early to judge.Checked Sep 29, 2026
  2. AI team

    Define your customer avatar

    Describe one named person who stands in for your best customer: their wants, pains, hangouts, price expectations and what they are not. Your AI team uses it when deciding product, price, voice and channels.

  3. Creative

    Ad copy only uses claims you've approved

    Your approved product claims now refresh from your live storefront, so ad copy only uses claims you have signed off, and product images default to each product's main storefront colour.

Earlier product · Dec 2025 – Aug 2026

Time-boxed autonomy grants with spend caps, and full decision provenance

  1. AI teamCouldn't tell · Too little data

    Time-limited autonomy with spend caps

    You can let an agent act without asking for a set period, with a spend cap on every action. When the period ends it goes back to asking for approval, so extra trust is a trial that ends on its own rather than a permanent switch.

    Expected result
    We expect founders to hand routine actions to agents sooner, and spend less time approving thembecause a grant that expires and is capped limits the downside of saying yes.
    How we'll know
    Share of routine agent actions carried out under a grant instead of waiting for approval, and how long approvals wait; success is shorter waits with no action ever exceeding its cap.
    What happened
    No time-limited grants have been issued yet, so every action still waits for approval and there is nothing to compare.Checked Sep 29, 2026
  2. AI team

    Revise a proposal with a note

    On any pending approval card you can write what should change, and the agent rewrites that same card instead of you rejecting it and waiting for a new proposal.

  3. Brand OS

    See what every agent decision was based on

    Each agent decision now shows the data and findings it was based on. Changes you make by hand in your store or email tool are logged separately, so results are credited to whoever actually made the change.

  4. Onboarding

    Add a brand and its collections in one step

    From settings, you can create a new brand together with its product collections in one form.

Agents hand off launch work to each other and plan ad budgets in safe steps

  1. Sales & leadsCouldn't tell · Too little data

    Ad budgets move in small, capped steps

    An agent now proposes budget changes as one small move at a time, shifting spend toward your best-performing ad sets. Every proposal stays within your daily spend cap.

    Expected result
    We expect a better return on ad spend at the same budgetbecause money moves gradually toward ad sets that are already converting instead of staying spread across winners and losers.
    How we'll know
    Return on ad spend in the weeks after budget moves compared with the weeks before, at similar total spend; success is a higher return with no day over the spend cap, disproved if it stays flat or falls.
    What happened
    Only one budget move has been made so far, for our single pilot brand, too little to compare return before and after.Checked Sep 29, 2026
  2. Brand OSCouldn't tell · Too little data

    Every collection runs as its own launch test

    Each product collection now has its own launch test with its metrics, evidence and forecast, a verdict on whether to launch, and a diagnosis of the funnel step where shoppers drop off and the likely reason.

    Expected result
    We expect faster, cheaper go or no-go calls on new collectionsbecause each one gets its own verdict and a named leak in the funnel instead of being judged on blended store numbers.
    How we'll know
    Time from a collection's first ad spend to a launch, extend or stop decision; success is a decision inside the test window for every collection, disproved if collections keep running past it without one.
    What happened
    This has run for one pilot brand only. Early signs are poor: collections that went live a month ago are still running without a launch, extend or stop decision, so we'll add a hard deadline that forces a decision.Checked Sep 29, 2026
  3. AI team

    Agents hand work to each other

    When the creative agent finishes a product pack and you approve it, the marketing agent picks it up to build the campaign. The founder agent reviews each deliverable before the next desk starts, and the Team view shows every handoff.

  4. Creative

    New creative stays inside your brand's look

    Brand OS builds a visual design profile from your storefront, shown in a Design system panel, and every new creative stays inside it. Product renders now start from your real product photos, so the product shown is your actual product.

  5. Brand OS

    Creative scoreboard and a marketing to-do list

    • A scoreboard ranks your ads by results within each launch test
    • A Marketing panel lists the approvals and inputs waiting on you

Full Meta ads sync, abandoned-cart recovery and a Today tab of decisions

  1. Sales & leadsCouldn't tell · Too little data

    Abandoned-cart recovery and approved email campaigns

    Brand OS spots shoppers who start a checkout and leave, and for those who gave an email it triggers your recovery flow in Klaviyo. Agents can also draft email campaigns; nothing sends until you approve, and results are reported back.

    Expected result
    We expect to win back a share of abandoned checkouts that were previously lostbecause shoppers who left an email now get a timely reminder.
    How we'll know
    Recovered orders as a share of abandoned checkouts with an email; success is a recovery rate that holds week over week once reminders are going out, disproved if recovered orders stay at zero after a month of sends.
    What happened
    Reminders haven't started going out yet, and almost none of the abandoned checkouts so far left an email, so recovery can't be measured. Capturing emails earlier in checkout is the first fix.Checked Sep 29, 2026
  2. AnalyticsCouldn't tell · Too little data

    Real contribution margin for every product

    Enter each product's unit cost, fulfilment and payment fees once. Profit and loss, scoreboards, experiments and agent decisions then all use margin after those costs instead of revenue.

    Expected result
    We expect fewer money-losing scale decisionsbecause founders and agents judge products on margin after costs, not on revenue or ad return alone.
    How we'll know
    Share of live products with unit costs entered, and share of scale decisions made on products with positive margin; success is costs entered for every live product within two weeks of setup.
    What happened
    Unit costs have been entered for only one product at our single pilot brand, too few to judge. Costs missing for most live products is itself the gap, so we'll ask for them during setup.Checked Sep 29, 2026
  3. Sales & leadsCouldn't tell · Not measured

    Your whole ad account inside Brand OS

    Every campaign, ad set and ad from Meta now syncs in with daily results and trend charts, plus a change history showing who changed what, so changes made by agents and by you can be told apart.

    Expected result
    We expect ad proposals to be grounded in real resultsbecause agents can now see every campaign's daily numbers and what changed, instead of a partial view.
    How we'll know
    Daily spend and results in Brand OS match the ad account's own reporting, and ad proposals cite them; success is a close daily match, disproved if the numbers keep drifting apart.
    What happened
    Daily ad numbers are syncing for our pilot brand, but we don't check them against the ad account's own reporting each day, so we can't confirm they match. We need a daily comparison before we can grade it.Checked Sep 29, 2026
  4. Brand OS

    A Today tab for decisions waiting on you

    Every approval an agent needs from you now sits in one queue on Today, each with a plain-language summary of what the agent wants to do and the evidence behind it.

  5. Brand OS

    Scale or stop alerts for each product

    Launch gates watch demand signals for each product and alert you once when it crosses a scale or stop threshold. The go or no-go decision then waits in your approvals queue.

  6. AI team

    An engineering agent and a Work tab

    An engineering agent keeps your store's development board in sync and flags failing checks and stalled work. A new Work tab gives each agent a desk showing what it is working on and what it did recently.

  7. Onboarding

    Setup checklist and storefront funnel

    A checklist walks you through connecting your store and tools, and a funnel view shows visits, sign-ups and checkout attempts step by step so you can see where shoppers drop off.

Your AI team now runs marketing, creative, support and campaigns end to end

  1. AI teamCouldn't tell · Too little data

    An AI team that works on a schedule

    Marketing, creative and customer support agents, plus a weekly founder review, now run on a schedule. Each reads live store, ad and support data and files proposals in an approval queue, and nothing that changes your store or ads happens until you approve it.

    Expected result
    We expect founders to spend less time on routine marketing, creative and support workbecause agents prepare the work and the founder only approves or rejects it.
    How we'll know
    Share of agent proposals approved, and time founders spend in the approval queue each week; the bar we set is more than half of proposals approved within the first month, disproved if most are rejected.
    What happened
    This has run for one pilot brand only, so we can't grade it yet. The early read is poor: only about one in twenty-five resolved proposals were approved and most were rejected or expired, so we're cutting low-value proposals and making each one easier to judge.Checked Sep 29, 2026
  2. Sales & leadsCouldn't tell · Not measured

    Agents build and launch ad campaigns

    The marketing agent assembles a complete Meta campaign from creative you've approved. It goes live only after you approve it, and every change is tracked.

    Expected result
    We expect new products to reach paid traffic soonerbecause building a full campaign no longer waits on the founder doing it by hand.
    How we'll know
    Time from approved creative to a live campaign; success is campaigns live within a week of creative approval, disproved if approved creative keeps sitting unused.
    What happened
    There's one pilot brand and we don't link an approved creative to the campaign that later uses it, so time to launch can't be measured. Early signs show approved images waiting weeks before an agent-built ad went live.Checked Sep 29, 2026
  3. CreativeCouldn't tell · Too little data

    A library of AI-made product images

    The creative agent turns briefs into on-brand product images and saves them to a creative library, where you review them before any is used in an ad.

    Expected result
    We expect brands to test more ad creative for lessbecause images come straight from a brief instead of waiting on a shoot or a designer.
    How we'll know
    Share of generated images approved for use in ads; success is more than half approved without a redo, disproved if most are rejected.
    What happened
    Only one pilot brand has used the library, and around one in ten images was approved for ads, with the rest retired. The sample is too small to grade, but it points to image quality needing work before this scales.Checked Sep 29, 2026
  4. Brand OS

    Brand thesis, monthly goals and weekly reviews

    Write down what your brand believes and set monthly goals. Agents plan their work against them, and a weekly review grades progress toward those goals.

  5. AI team

    Connect your tools and track experiments

    • Connect the other tools your brand uses so agents can work with them
    • Follow the experiments agents run, each graded against real outcomes
  6. AI team

    Decision log and earned autonomy

    A decision log and a page per agent show what each agent decided, why, and whether it paid off. Agents earn more autonomy for a type of action as their track record on it improves.

Pre-order tiers with batch sizes for demand-validation launches

  1. Sales & leadsCouldn't tell · Too little data

    Pre-order tiers with a set batch size

    Shoppers can reserve a product with a deposit in one of two tiers, each capped at a set batch size, so early reservations fill a defined production run. The Pre-order panel shows how many spots each tier has taken.

    Expected result
    We expect a clearer read on demand before committing to productionbecause paid deposits against a fixed batch are a stronger buying signal than sign-ups.
    How we'll know
    Share of batch spots reserved with a deposit during the launch test; success is the batch filling within the test window, disproved if deposits stay well short of it.
    What happened
    Our one pilot launch has taken a single deposit so far, too little to read demand against a batch.Checked Sep 29, 2026

Brand OS goes live: connect Shopify, Meta, GA4 and Klaviyo and track pre-orders

  1. Brand OSCouldn't tell · Not measured

    Connect your store, ads, analytics and email for each brand

    In each brand's settings, connect Shopify, Meta, Google Analytics 4 and Klaviyo, and the Brand OS dashboard fills with live numbers: site traffic, funnel steps, ad performance and email list size.

    Expected result
    We expect founders to start their day in one dashboard instead of four separate toolsbecause store, ad, analytics and email numbers now sit side by side for each brand.
    How we'll know
    Share of brands with all four connected that open the dashboard on most days; the bar we set is a majority within four weeks, disproved if daily use stays rare.
    What happened
    Only one pilot brand has all four connected, and we don't record daily dashboard visits, so daily use can't be measured. We need to record visits before we can grade it.Checked Sep 29, 2026
  2. Brand OSCouldn't tell · Not measured

    Pre-order command center

    One panel tracks a pre-order launch: signups today and over the last 7 days, deposits and deposit revenue, email list growth, a daily signup trend and how many founding spots have been claimed.

    Expected result
    We expect founders to spot a stalling launch early and adjust ads or offers soonerbecause signups, deposits and list growth are visible in one place every day.
    How we'll know
    Time from a sustained dip in signups or deposits to a change in ads or offer; the bar we set is a response within two days, disproved if dips still go unnoticed for a week.
    What happened
    We run this with a single pilot brand and don't record when a founder reacts to a dip, so response time can't be measured. We'll log dips and the changes that follow.Checked Sep 29, 2026
  3. Brand OS

    Run several brands from one portfolio

    Create several brands in Brand OS and edit each one's details in its own settings, so one account can oversee a whole portfolio.

Pre-order launches: waitlist, referrals and deposits tracked back to Meta ads

  1. Brand OSCouldn't tell · Too little data

    Pre-order launches with waitlist, referrals and deposits

    Brands can open a waitlist where every signup gets a referral link to move up the queue, and take paid founding-member deposits through their store. Each deposit is reported back to Meta as a purchase, so campaigns learn to find people who actually pay, not just people who click.

    Expected result
    We expect a lower ad cost per depositbecause Meta campaigns now optimise toward real deposits instead of clicks or signups, and referrals add signups at no ad cost.
    How we'll know
    Meta cost per founding-member deposit and share of waitlist signups that come from referrals; the bar we set is cost per deposit falling over the first weeks of optimisation, disproved if it stays flat.
    What happened
    Our one pilot launch drew only a single deposit, too few for a cost-per-deposit trend. None of its waitlist sign-ups came through a referral so far, so we'll revisit how referrals are offered.Checked Sep 29, 2026
  2. Brand OS

    Lead ads and newsletter signups flow into your waitlist

    People who fill in a Meta lead ad form or sign up for your newsletter are added to your waitlist and your Klaviyo list automatically, with no manual export.

First look at Brand OS: one dashboard for a D2C brand's store, ads and P&L

  1. Brand OSCouldn't tell · Not measured

    First preview of Brand OS

    A new dashboard for running a D2C brand puts the store, marketing, P&L, pricing and inventory, customer support, items waiting for your approval, a decision log and a 7-day summary on one screen. This first preview ran on sample data while live connections were built.

    Expected result
    We expect founders to run day-to-day brand decisions from one screenbecause every part of the business and everything waiting on them is gathered in one place.
    How we'll know
    Share of pilot brands that use the dashboard as their daily starting point once live data is connected; the bar we set is a majority, disproved if they keep working from separate tools.
    What happened
    We have one pilot brand with live data and don't record dashboard visits, so daily use can't be measured. We need to record visits before we can grade it.Checked Sep 29, 2026

Brand Factory: turn a product catalog into a launch-ready brand

  1. Brand OSCouldn't tell · Too little data

    Brand Factory: from product catalog to brand kit

    Upload a product catalog and AI reads the products, identifies likely target audiences, then drafts seven brand documents: customer persona, positioning, brand strategy, messaging, visual identity, packaging direction and a content growth playbook.

    Expected result
    We expect founders launching a new brand to reach a usable brand kit much fasterbecause the audience research and first drafts are generated from the catalog they already have.
    How we'll know
    Share of brand kits whose documents are kept with light edits rather than rewritten; the bar we set is most of the seven documents kept, disproved if founders rewrite most of them.
    What happened
    No brand kits have been saved yet, so we can't measure how many documents were kept. We'll measure it once founders start building kits.Checked Sep 29, 2026

Small brand profile improvements

  • Content & calendarSmall improvements to the marketing messages section of the brand profile.

Persona-based problem-and-solution stories for LinkedIn Company Pages

  1. Content & calendarCouldn't tell · Not measured

    Problem-and-solution LinkedIn stories for each buyer persona

    Add the buyer roles you sell to on your profile, such as a facility manager or IT administrator, and LinkedIn Company Page calendars plan story posts around one role's day-to-day problem and a real product feature that solves it.

    Expected result
    We expect B2B Company Page posts to feel more relevant to buyersbecause each story is built on a specific role's problem and a real feature instead of generic pain points.
    How we'll know
    Engagement rate on problem-and-solution stories compared with other Company Page posts; the bar we set is higher engagement within six weeks, disproved if they perform the same or worse.
    What happened
    Only a couple of these stories were published and we don't record engagement per post, so there's no comparison yet.Checked Sep 29, 2026
  2. Content & calendar

    Better LinkedIn post visuals

    LinkedIn feature spotlights now use the actual product features on your profile, visuals follow your design profile more closely, the LinkedIn preview looks more like the real post, and you can regenerate your design profile to refresh the visual style.

  3. Social publishing

    Editable, more relevant LinkedIn hashtags

    • Edit LinkedIn hashtags right under the caption in the post preview
    • Suggested hashtags are now specific to your brand instead of generic defaults

Product showcase posts and videos made from your own product photos

  1. VideoDidn't work

    Product showcase posts and videos from your photos

    Upload your product photos and Space AI builds showcase posts and multi-shot videos around them, with several product video styles to choose from. Every result is checked automatically so the product looks right before it reaches you.

    Expected result
    We expect product brands to publish more product-led contentbecause showcase posts and videos now feature their real products instead of generic imagery.
    How we'll know
    Share of product showcase posts and videos published without regeneration; the bar we set is a majority within four weeks, disproved if most are regenerated or discarded.
    What happened
    None of the product showcase posts or videos were published, and most never got past the idea stage. We'll pause adding showcase formats until we know why brands pass on them.Checked Sep 29, 2026
  2. Content & calendarDidn't work

    Fill calendar gaps and regenerate single ideas

    One click generates ideas for any empty dates between today and your last scheduled post, and a refresh button regenerates a single idea in the same slot without touching the rest of the calendar.

    Expected result
    We expect fewer gaps in brands' posting schedulesbecause empty dates can now be filled in one click instead of adding ideas by hand.
    How we'll know
    Share of active brands with no empty posting days in the next two weeks; the bar we set is a clear rise within a month, disproved if gaps persist.
    What happened
    Gaps grew instead of shrinking: active calendars covered about six of the next 14 days in May and about three by late June, and almost none were fully filled. One-click filling wasn't enough to keep calendars topped up, so we'll look at refilling calendars automatically.Checked Sep 29, 2026
  3. Content & calendar

    Captions checked against your company rules

    • Every generated caption is checked against the company-specific rules on your profile
    • Your brand name is always written exactly the way you spell it
  4. Content & calendar

    LinkedIn visuals match your Instagram style

    LinkedIn visuals are now made with the same design system as Instagram posts, so both channels look consistent, and the calendar shows when a post is queued, still processing or has stalled.

AI video reels from your calendar, plus a prepaid wallet for Meta ads

  1. Sales & leadsCouldn't tell · Too little data

    Prepaid ad wallet for Meta campaigns

    Top up an ad wallet by card and fund your Meta campaigns from it, with your balance, transaction history and active campaigns in one place. A campaign only goes live when the wallet can cover it, and Meta spend is capped at your balance.

    Expected result
    We expect brands to hand their Meta ad spend to Space AI with confidencebecause a campaign can't go live unfunded and spend can never run past what they topped up.
    How we'll know
    Spend beyond the funded balance, and share of ad-management brands that fund a campaign from the wallet; the bar we set is zero overspend and most active ad brands topping up within a month, disproved by any overspend or little take-up.
    What happened
    Only one brand has set up the wallet so far, too few to judge take-up. No overspend has been recorded.Checked Sep 29, 2026
  2. VideoCouldn't tell · Too little data

    AI video reels from calendar ideas

    Pick a video idea in your content calendar and get a finished reel: AI-generated clips stitched together with captions and your logo, checked automatically for visual problems before it reaches you.

    Expected result
    We expect brands to post video more oftenbecause a reel now comes straight from an idea already in their calendar, with no filming or editing.
    How we'll know
    Share of generated reels that are published; the bar we set is at least half within four weeks, disproved if most are discarded or regenerated.
    What happened
    Only a couple of reels were generated from calendar ideas and neither was published, too few to judge. We also need to track downloads, since some reels may be posted by hand.Checked Sep 29, 2026
  3. Content & calendar

    Separate founder and Company Page LinkedIn plans

    LinkedIn plans now treat a founder's personal profile and the Company Page as different accounts, with post types suited to each, including a new industry thought leadership format for founders.

Create a digital twin that presents videos for you

  1. VideoCouldn't tell · Too little data

    Your own digital twin for talking-head videos

    Upload a short training video and a consent recording to create an AI avatar of yourself, then use it to present talking-head videos for ideas in your content calendar without filming each one.

    Expected result
    We expect founders to publish more face-to-camera videobecause they record once and the twin presents every video after that.
    How we'll know
    Share of brands that create a twin and publish at least two twin videos within a month; the bar we set is a majority, disproved if most stop after the first video.
    What happened
    No brand has created a digital twin, so there's nothing to measure. The consent and recording step may be too heavy, and we'll check that with brands before building more.Checked Sep 29, 2026
  2. Content & calendar

    Edit directly from a shared calendar preview

    Anyone with a shared calendar preview link and the account password can change ideas, captions and images right in the preview, without signing in to the dashboard.

Creatives are auto-checked for typos and your brand rules before they're saved

  1. CreativeDidn't work

    Automatic typo and brand-rule checks on creatives

    Every generated creative is checked for spelling mistakes and against custom rules you add on your brand profile, such as required labels or how your logo is shown. Anything wrong is repaired automatically before the creative is saved, and the calendar shows each post's check status.

    Expected result
    We expect fewer creatives to need manual fixes before publishingbecause typos and broken brand rules are caught and repaired before anyone sees them.
    How we'll know
    Share of generated creatives approved without edits; the bar we set is a clear rise over the month before, disproved if edit rates don't change.
    What happened
    Manual image fixes didn't fall after the checks shipped. The share of posts edited by hand rose from about 2% in May to about 5% in June, so we'll make sure the checks fix the issues brands actually edit for.Checked Sep 29, 2026
  2. Content & calendar

    Schedule any post for a time you choose

    Pick a date and time for any generated post and it publishes then, instead of waiting for the next slot in your posting schedule.

  3. Content & calendar

    One-off posts that follow your strategy

    When you create a one-off post you choose the platform and content type, now including LinkedIn, and it's written with your strategy's tone, audience and content pillars.

  4. Content & calendarCouldn't tell · Not measured

    Calendar ideas checked against their format

    Each idea in a generated calendar is checked against its post format, so a carousel idea really reads as a carousel, and ideas that don't fit are rewritten before you see them.

    Expected result
    We expect calendars to need fewer manual rewrites, so an account can run with little hands-on editingbecause ideas that don't fit their format are fixed before they reach the calendar.
    How we'll know
    Share of calendar ideas edited or regenerated by hand before publishing; the bar we set is a clear drop versus before the checks, disproved if it doesn't move.
    What happened
    We don't record when an idea is edited by hand or regenerated, so we can't compare before and after. We need to record that before we can grade it.Checked Sep 29, 2026

Publish to Instagram and Facebook together in one step

  1. Social publishingCouldn't tell · Not measured

    Publish to Instagram and Facebook in one step

    Choose Instagram and Facebook together when you publish and the post goes out to both at once, with a link to each live post afterwards.

    Expected result
    We expect more posts to reach Facebook as well as Instagrambecause posting to both now takes the same single step as posting to one.
    How we'll know
    Share of Instagram posts also published to Facebook by brands with both connected; the bar we set is a majority within four weeks, disproved if it stays low.
    What happened
    Posting to Facebook was switched on for every published Instagram post, but we don't record whether the Facebook copy actually went live, so we can't confirm it. We need to record Facebook publishes before we can grade it.Checked Sep 29, 2026
  2. Plans & billing

    Self-serve card checkout and billing portal

    • Subscribing now goes through a secure hosted card checkout
    • A billing portal lets you update your card, change plans or cancel without contacting us
  3. Plans & billing

    Pick only the modules you need

    The redesigned pricing page lets you build a plan from individual modules or pick a ready-made bundle, with a checkout summary before you pay.

  4. Content & calendar

    One design engine for all carousels

    Instagram carousels and LinkedIn document carousels are now made by the same design engine, so slides look consistent and polished on either platform.

LinkedIn polls and document posts come to the content calendar

  1. Social publishing

    LinkedIn polls and document posts in the calendar

    Generated calendars can now include LinkedIn poll ideas with the question and answer options already written, and LinkedIn document posts that read as swipeable slides.

  2. Creative

    Labelled reference images for image edits

    When you ask for changes to a post image, you can attach reference images and label what each one is for, so the edit uses every image the way you intended.

Early performance marketing workspace with ad budget recommendations

  1. Sales & leads

    Ad budget recommendations across channels

    A new performance marketing workspace recommends how to split an ad budget across channels, with a forecast for each channel, industry benchmarks in your currency and the reasoning behind every recommendation. An early A/B tests view sits alongside it.

  2. OnboardingCouldn't tell · Not measured

    More accurate brand colors and logos from your website

    During setup, brand colors are now chosen by comparing your website's design, styling and logo instead of trusting a single source, and logos are found on more kinds of site layouts.

    Expected result
    We expect fewer brands to correct their colors or logo after setupbecause colors are now agreed across several signals from the website instead of one guess.
    How we'll know
    Share of new brands that change their brand colors or logo in the first week after setup; the bar we set is a clear drop versus the month before, disproved if it doesn't fall.
    What happened
    We don't keep a history of brand color or logo changes after setup, so we can't compare correction rates before and after. We need a record of those changes before we can grade it.Checked Sep 29, 2026

SEO blog posts, AI video studio and LinkedIn company page publishing

  1. Content & calendarMixed

    SEO blog posts from your content strategy

    Add Web as a platform in your marketing strategy and the calendar plans blog posts written for search, which you can generate and download to publish on your own site.

    Expected result
    We expect brands to use Space AI for website content as well as socialbecause blog posts now come from the same strategy that already plans their social calendar.
    How we'll know
    Share of active brands that add Web as a platform and download at least one blog post within four weeks; the bar we set is one in five, disproved if almost none do.
    What happened
    About one in seven active brands added blog posts to their calendar, meaningful but short of our one-in-five bar. We don't record blog downloads, so we can't confirm how many were actually used; we need to record them before investing further.Checked Sep 29, 2026
  2. VideoCouldn't tell · Too little data

    Video studio: from idea to finished reel

    Start from an idea and get a storyboard, then generate each scene as an AI video clip or have an AI presenter deliver it. The clips are stitched into one finished reel.

    Expected result
    We expect brands to publish more videobecause a reel no longer needs filming or editing, only an idea.
    How we'll know
    Share of reels finished in the studio that go on to be published; the bar we set is at least half within four weeks, disproved if most are abandoned.
    What happened
    Only one studio reel was started and it was never finished, so there's no publish rate to measure. Take-up, not finishing, was the problem, and we'll look at why before building more video tooling.Checked Sep 29, 2026
  3. Social publishingCouldn't tell · Too little data

    Publish to your LinkedIn company page

    Connect your LinkedIn company page and publish to it directly, not only to a personal profile. The calendar can also plan LinkedIn-specific formats: polls, document carousels and product feature posts built from the features on your business profile.

    Expected result
    We expect more B2B brands to keep a steady LinkedIn presence through Space AIbecause they can now post as the company, where their buyers follow them, and not only as a person.
    How we'll know
    Share of brands with a company page connected that publish to it at least weekly; the bar we set is a majority within four weeks of connecting, disproved if company-page posts stay rare.
    What happened
    Only one or two brands ever connected a LinkedIn company page, too few to judge weekly posting. Connecting LinkedIn at all is the first hurdle to fix.Checked Sep 29, 2026
  4. Creative

    Describe a post and get the right format

    Type what you want to post in plain words (for example, "we hosted a hackathon") and the right post type is chosen for you, including banners and product shoots generated at the correct sizes.

  5. Content & calendar

    Competitor research feeds your content strategy

    • Add your competitors and Space AI researches what they post
    • Those findings feed into your strategy and the ideas in your calendar
  6. Content & calendar

    Faster calendars and simpler competitor inputs

    Content calendars are generated noticeably faster, and the competitor section is now split into separate fields for keywords, competitor websites and target personas.

Plan posts for Facebook, X, YouTube and Reddit, plus AI post scoring

  1. Social publishingDidn't work

    Plan and publish for Facebook, X, YouTube and Reddit

    Connect Facebook, X, YouTube and Reddit and your content calendar plans and publishes posts for them alongside Instagram and LinkedIn, so one calendar covers every channel.

    Expected result
    We expect more brands to run all their social channels from one calendarbecause they no longer need a separate tool for platforms beyond Instagram and LinkedIn.
    How we'll know
    Share of active brands publishing to at least one of the four new platforms; the bar we set is one in five within six weeks, disproved if almost none connect them.
    What happened
    About one in seven active brands got ideas for these platforms, but none published to them through us. We'll concentrate on the channels brands actually publish to rather than widening platform coverage.Checked Sep 29, 2026
  2. Creative

    AI scoring for every post

    Get an AI score for any Instagram or LinkedIn post, with specific suggestions to improve it that you can apply in one click instead of rewriting the post yourself.

  3. Creative

    Edit post images in plain words

    Describe the change you want to a post's image and an edited version is generated in the background, ready for you to review.

  4. Onboarding

    Chat to edit your landing page

    • Change your landing page by describing edits in a chat with AI
    • Upload your own images to use on it
  5. Content & calendar

    Edit content ideas inline

    • Change an idea's details directly in the calendar without opening a separate editor
    • Add new ideas through a redesigned form

Brand builder and AI reel editor: build a brand kit and turn raw footage into reels

  1. OnboardingCouldn't tell · Too little data

    Brand builder for new businesses

    Describe your business and a guided builder creates a logo, brand colors and messaging, then puts up a simple public landing page you can share.

    Expected result
    We expect to win businesses that don't yet have a brand identity or websitebecause they can get a usable brand kit and a page to share in one session, then go straight to making content.
    How we'll know
    Share of brand-builder starts that end with a published landing page, and how many of those go on to generate a content calendar; success is most starts finishing.
    What happened
    Only a few brand-builder sessions were ever started, too few to judge completion or follow-on calendar use. Most new sign-ups already had a brand, so demand for this path looks small.Checked Sep 29, 2026
  2. VideoCouldn't tell · Too little data

    Reel editor for raw footage

    Upload raw talking-head video and get back an edited reel: speech is transcribed into captions, framing follows the speaker's face, audio is cleaned up and your logo and colors are applied.

    Expected result
    We expect brands to publish short video without hiring an editorbecause a phone recording becomes a finished, branded reel automatically.
    How we'll know
    Share of uploaded videos that produce a reel the brand downloads or publishes; success is above half, disproved if most reels are abandoned.
    What happened
    We have no recorded uploads of raw footage, so there is nothing to measure, and brands didn't pick this up. We also don't record reel downloads, which we'd need to measure it if usage returns.Checked Sep 29, 2026
  3. Onboarding

    Update your profile from documents

    Upload business or brand documents and Space AI pulls out profile details. You see what was found and choose which fields to apply.

  4. Creative

    Every post designed by AI

    The old fixed border templates are gone. Every post is now designed by AI around its own idea and your brand look, instead of reusing the same few frames.

Product showcase posts from your own photos, plus festival-aware calendars

  1. CreativeDidn't work

    Product showcase posts from your photos

    Attach photos of your products to an idea and AI places them in on-brand creative, so the post shows the real product rather than a generic stand-in.

    Expected result
    We expect product brands to publish more of their generated postsbecause the creative features the actual product a customer can buy.
    How we'll know
    Publish rate of product showcase posts versus other generated posts for product brands; success is a higher publish rate within four weeks.
    What happened
    None of the product showcase ideas were published, while about 4% of other generated ideas were. Few showcase ideas even got as far as a finished post, so we'll find out why brands pass on them before making more.Checked Sep 29, 2026
  2. Content & calendar

    Holiday posts matched to your country and industry

    The calendar now suggests holiday and festival posts based on your country and industry, covering major Indian festivals, global holidays and industry days.

  3. Content & calendar

    Share your calendar with a link

    Create a read-only link to your content calendar so a teammate or client can review planned posts without an account.

  4. Social publishing

    More LinkedIn post formats and a branding toggle

    • Calendars now plan posts in formats and styles made for LinkedIn
    • Each post has a switch to show or hide your branding details
  5. Creative

    Spotlight posts for people

    Feature a team member or customer with their photo and story in a designed spotlight post.

Analytics adds competitor benchmarks, an AI analyst and downloadable reports

  1. AnalyticsCouldn't tell · Not measured

    Competitor benchmarks and posting insights

    Analytics now has tabs for audience, posting patterns, competitors and reports. Compare your accounts with competitors, see your best time to post and which content themes work, track a posting consistency score and get a predicted performance for posts.

    Expected result
    We expect brands to get more engagement per postbecause they can see when to post and which themes work, and to value Space AI more because they can see how they compare with competitors.
    How we'll know
    Weekly share of active brands that open Analytics, and engagement on posts published at the suggested best time versus other times; success is higher engagement at suggested times.
    What happened
    We don't record Analytics visits or engagement on each published post against the suggested time, so this can't be measured yet. We need to record both before we can grade it.Checked Sep 29, 2026
  2. Creative

    Edit post images by chatting

    Describe the change you want in plain words and AI edits the post image. Save the result as the post's new image.

  3. Analytics

    AI analyst and PDF reports

    Ask questions about your numbers in plain language and get answers from your own data. Export a branded PDF report and pick its sections, such as an executive summary, top posts and a competitor benchmark.

  4. OnboardingDidn't work

    Guided onboarding with a live preview

    Setup walks you through adding your logo and design images, shows a live preview of your posts and offers an interactive demo. Afterwards, a checklist takes you through creating a strategy, generating a calendar, connecting Instagram and publishing your first post.

    Expected result
    We expect more new brands to publish their first postbecause the checklist lays out the four steps that get them there.
    How we'll know
    Share of new sign-ups who publish a first post within seven days; success is a clear rise over the month before.
    What happened
    Fewer than one in a hundred new sign-ups published a first post within seven days, both before and after the checklist, so there was no rise. Most sign-ups never connected a social account, which is the step we'll now make easier.Checked Sep 29, 2026
  5. Analytics

    Time and money saved

    Analytics shows an estimate of the time and money Space AI has saved you on content work.

Campaign strategies that steer your calendar, plus Analytics and Instagram auto-replies

  1. Content & calendarCouldn't tell · Not measured

    Campaign strategy steers your calendar

    Set your objective, audience, competitors, platforms and posting schedule in a campaign strategy, or let AI draft it. Your content calendar and each post are then planned around that strategy.

    Expected result
    We expect brands to keep and publish more of the ideas we suggestbecause the calendar is planned around what the brand is trying to achieve rather than generic ideas.
    How we'll know
    Share of calendar ideas that brands keep rather than delete or regenerate, for brands with a strategy versus without; success is a higher keep rate with a strategy.
    What happened
    Almost every calendar was built with a strategy, so there's no group without one to compare, and we don't record when ideas are deleted or regenerated. We need to record that before we can grade it.Checked Sep 29, 2026
  2. Content & calendarCouldn't tell · Too little data

    Food and beverage mode

    Restaurants, cafes, bars and bakeries get a dedicated mode. Add photos of dishes or your outlet and AI turns them into polished promo posts on a food-focused calendar, which you can export as a grid PDF.

    Expected result
    We expect to sign and keep more food and beverage businessesbecause their content depends on real dish and outlet photos that generic generated imagery can't replace.
    How we'll know
    Share of food and beverage sign-ups still publishing after four weeks; success is retention at or above other industries.
    What happened
    Only a handful of food and beverage businesses signed up, and very few brands in any industry publish through us, so there's no fair retention comparison yet.Checked Sep 29, 2026
  3. Social publishingCouldn't tell · Too little data

    Instagram auto-replies in your brand voice

    Space AI answers Instagram DMs and comments in your brand voice. When a message looks like a complaint, auto-replies pause for that conversation and it goes to a person until they mark it resolved. A test page lets you try replies before going live.

    Expected result
    We expect brands to answer customers faster and lose fewer sales to unanswered messagesbecause common questions get an on-brand reply right away while complaints still reach a person.
    How we'll know
    Share of incoming DMs and comments answered automatically without escalation, and time to first reply; success is most messages answered with no one needed and every complaint escalated.
    What happened
    No brand has turned on auto-replies, so there is nothing to measure. We'll find out whether brands don't trust automated replies or simply haven't found the setting.Checked Sep 29, 2026
  4. AnalyticsCouldn't tell · Not measured

    Analytics for Instagram and LinkedIn

    A new Analytics tab brings in your Instagram and LinkedIn numbers: reach, engagement, follower growth and how each post performed. It adds a KPI scorecard, industry benchmarks, audience demographics and which content types work best for you.

    Expected result
    We expect brands to post more of what works and to see Space AI's value in their own numbersbecause results now sit next to the posts that produced them.
    How we'll know
    Share of active brands that open Analytics at least weekly; success is a majority within a month, with engagement per post rising for brands that use it.
    What happened
    We don't record Analytics page visits or per-post engagement over time, so weekly use and its effect can't be measured. We need to record both before we can grade it.Checked Sep 29, 2026
  5. Creative

    Finished post designs with text built in

    Posts are now generated as complete designs with the text built into the image, guided by a design profile of your style, colors, typography and mood taken from your logo and brand guidelines. Each design is checked for clipped or hard-to-read text.

  6. Creative

    Assets tab for images and your own templates

    Keep your images in one library and upload your own post templates for new posts to use. You can also set text box backgrounds and position text vertically.

  7. Content & calendar

    Captions written for each platform

    Instagram captions now include search keywords, hashtags and image alt text, while LinkedIn captions stay clean and professional.

Publish to LinkedIn and turn your own photos into branded posts

  1. Social publishingDidn't work

    Publish to LinkedIn

    Connect a LinkedIn account and publish to it from Space AI, including text-led thought-leadership posts you create in the content calendar.

    Expected result
    We expect Space AI to serve more service and business-to-business brands, and brands to publish more each weekbecause LinkedIn is where many of them reach customers and it was not supported before.
    How we'll know
    Share of active brands that connect LinkedIn and publish at least one post there within a month; success is one in four or more, disproved if it stays under one in ten.
    What happened
    Most active brands got LinkedIn ideas in their calendar, but only about 2% published to LinkedIn through us, far under our one-in-ten floor. Very few brands connected a LinkedIn account, so we'll make connecting part of setup instead of a separate step.Checked Sep 29, 2026
  2. CreativeDidn't work

    Turn your own photos into posts

    Upload your own photos when adding an idea and Space AI turns them into on-brand posts. Crop and fit settings are saved with the post, and downloads come out exactly as edited.

    Expected result
    We expect more generated posts to be approved and publishedbecause posts built on a brand's real products, people and places look more authentic than fully generated imagery.
    How we'll know
    Publish rate of posts made from uploaded photos versus fully generated posts; success is a higher publish rate for uploaded-photo posts within four weeks.
    What happened
    None of the posts built from uploaded photos were published, against roughly 4% of fully generated posts, although the sample was small. Brands uploaded photos but didn't take the result live, so we'll look at why these posts stall at review before investing further.Checked Sep 29, 2026
  3. Social publishing

    Schedule to Instagram, LinkedIn or both

    Choose the platforms for each post in its preview and it publishes to each one at the scheduled time. Text-only posts can be scheduled for LinkedIn.

  4. Plans & billing

    Plans activate on payment, with watermark styles

    A paid plan switches on as soon as the payment goes through. Paid plans can also choose a watermark style or remove the watermark.

Free, Premium and Enterprise plans with in-app checkout

  1. Plans & billingDidn't work

    Free, Premium and Enterprise plans

    Pick a plan and compare what each includes, upgrade and pay inside the app, and see your payment history. Each plan comes with a usage balance for AI generation.

    Expected result
    We expect Space AI to start earning recurring revenue from active brandsbecause brands can now upgrade on their own the moment they need more.
    How we'll know
    Share of active free brands that upgrade to a paid plan within 30 days; the bar is at least one in twenty, disproved if almost no one upgrades without talking to us.
    What happened
    About 2% of free brands moved to a paid self-serve plan, under our one-in-twenty bar; paying customers mostly came through a conversation with us. We'll treat sales-assisted onboarding as the main path to revenue and test what makes self-serve upgrading worthwhile.Checked Sep 29, 2026
  2. Content & calendar

    Sharper ideas from a deeper read of your website

    Space AI now reads more of your website when planning content, so post ideas and marketing messages reflect your actual products, services and positioning.

Sharper profile autofill and posts that look the same everywhere

  1. OnboardingCouldn't tell · Not measured

    More accurate profile autofill

    The AI that drafts your business profile now gets more details right, reads uploaded documents better, picks up fonts from brand documents and pulls your logo from your website.

    Expected result
    We expect new brands to finish setup faster with fewer correctionsbecause more of the profile is filled in correctly on the first pass.
    How we'll know
    Share of autofilled profile fields that new brands save without editing; success is a majority saved unchanged.
    What happened
    We don't record which autofilled profile fields a brand changes before saving, so we can't tell how many were kept as drafted. We need to record that before we can grade it.Checked Sep 29, 2026
  2. Content & calendarCouldn't tell · Not measured

    Posts look the same everywhere

    The preview you approve, the file you download and the post that goes live are now drawn the same way, so text, colors and the watermark no longer shift between them. Carousel PDFs use the same rendering.

    Expected result
    We expect fewer posts to be re-edited or taken down after publishingbecause what a brand approves in the preview is exactly what goes live.
    How we'll know
    Reports of a published or downloaded post not matching its preview; success is none in the month after release.
    What happened
    We have no record of preview mismatch reports either way, so we can't confirm none happened. We need a way to record those reports before we can grade it.Checked Sep 29, 2026

AI fills in your business profile from your website, Instagram or documents

  1. OnboardingCouldn't tell · Not measured

    AI drafts your business profile

    Paste your website or Instagram link, or upload a document, and AI reads it and fills in your business profile. Nothing is saved until you review the draft and confirm it.

    Expected result
    We expect more sign-ups to complete setup and reach their first content calendarbecause filling in a long profile form by hand is replaced by reviewing a draft.
    How we'll know
    Share of new sign-ups who finish their business profile on the day they join; success is a clear rise over the manual-form baseline, disproved if completion stays flat.
    What happened
    We have no manual-form baseline to compare against, because sign-up records start the same week this shipped, and we don't record when a profile counts as finished. We need to record when a profile is finished before onboarding changes can be measured.Checked Sep 29, 2026
  2. Social publishingWorked

    Schedule posts from the post preview

    Pick a date and time in the post preview and the post is queued. It publishes to Instagram automatically at that time, with no need to be online.

    Expected result
    We expect brands to publish more often and more consistentlybecause they can queue a week of posts in one sitting instead of posting each one live.
    How we'll know
    Posts published per active brand per week, and the share of them that were scheduled; success is scheduled posts becoming the majority of what brands publish.
    What happened
    About six in ten Instagram posts published through Space AI went out at a pre-set scheduled time, making scheduling the main way brands publish. Brands that publish do so a little more than once a week on average, but only a handful of brands publish at all.Checked Sep 29, 2026

A new Space AI website and more reliable Instagram publishing

  • Sales & leadsA redesigned Space AI website shows features, real customer posts and plans at a glance.
  • Content & calendarGenerated posts can carry an optional "Made by Space AI" watermark.
  • Social publishingInstagram connection and publishing are more reliable.
  • Speed & reliabilityPost images keep loading reliably long after they are created, plus other speed and reliability fixes.

Scheduled posts now publish automatically at the time you set

  1. Social publishingDidn't work

    Scheduled posts publish on their own

    Posts you schedule now go out automatically at their set date and time. Before publishing, Space AI checks that the account is still connected, so nobody has to be online to post.

    Expected result
    We expect brands to publish more consistently and spend less time posting by handbecause a post no longer needs someone to be around at the right moment to go live.
    How we'll know
    Share of scheduled posts that go live at their set time without anyone stepping in; the bar is 95% or more, and below 90% means it is not reliable enough to depend on.
    What happened
    About two in three scheduled posts went live within 30 minutes of their set time, well under our 90% floor. Instagram posts were on time about nine times in ten, but LinkedIn posts went out in delayed batches days late, so we are fixing LinkedIn scheduling and will alert when a post misses its slot.Checked Sep 29, 2026

Edit post text and colors directly on the design before you publish

  1. Creative

    Edit posts before you publish

    Open any generated post in an editor to change its wording, colors and template, then keep that as the final version. Small fixes no longer mean generating a new post.

  2. Social publishingWorked

    More reliable Instagram publishing

    Single-image and carousel posts now upload and publish to Instagram reliably. You get a confirmation once the post is live, and a clear message if it fails.

    Expected result
    We expect far fewer failed Instagram postsbecause post images are now prepared and stored where Instagram can fetch them before publishing starts.
    How we'll know
    Share of Instagram publish attempts that succeed on the first try; the bar is 95% or more, and under 90% means the fix did not hold.
    What happened
    About 97% of recorded Instagram publish attempts succeeded, above our 95% bar. We don't yet record retries separately, so this counts attempts rather than strictly first tries.Checked Sep 29, 2026

Publish posts straight to Instagram from your content calendar

  1. Social publishingWorked

    Publish to Instagram from the content calendar

    Connect your Instagram account once, then publish single-image and carousel posts straight from the calendar, without downloading images and uploading them in the Instagram app.

    Expected result
    We expect more generated posts to actually go livebecause publishing becomes one click instead of a download-and-upload step on a phone.
    How we'll know
    Posts published from Space AI per Instagram-connected brand per week; success is at least one a week in the first month, disproved if most connected brands still only download.
    What happened
    Four of the five brands that connected Instagram published from Space AI in most weeks, averaging well over one post a week while active. The catch is reach: only a handful of brands connected Instagram at all, so connecting is now the step to make easier.Checked Sep 29, 2026
  2. Content & calendar

    Rate and comment on generated posts

    Give any post a rating and a comment on what to change. Feedback is saved with the post so it can be used to make future posts closer to what you want.

Download carousels as PDFs, plus a new account profile page

  1. Creative

    Download carousels as PDFs

    Save a whole carousel as one PDF from the post preview, ready to upload as a document post.

  2. Creative

    Testimonial posts

    Add a customer testimonial and get it back as a designed social post in your brand style, with a layout built for quotes.

  3. Onboarding

    Profile page with self-serve account controls

    A new profile page lets you verify your email, change your email address and delete your account yourself, without contacting support.

Festive posts for Diwali, Christmas and more, built into your content calendar

  1. Content & calendarDidn't work

    Festive posts in your content calendar

    Your calendar now knows the upcoming festivals and holidays for your region and industry, such as Diwali and Christmas, and suggests posts for them. Those posts come with festive designs and imagery, ready to review.

    Expected result
    We expect more brands to post for the festivals and holidays that matter to their customersbecause ideas and designs now arrive in the calendar ahead of each date instead of being planned by hand.
    How we'll know
    Share of active brands that publish at least one festive post in the week of a major festival; success is a majority of active brands, disproved if festive ideas are mostly skipped.
    What happened
    Almost every active brand received festive ideas, but only about one in twenty of those brands published a festive post through us, and about 5% of festive ideas went live. The ideas weren't the bottleneck; getting brands connected and publishing was, so that is where we will focus.Checked Sep 29, 2026
  2. Onboarding

    Faster business setup

    Pick your industry from a list and start from a sample profile that fills in your business details for you. You can also point Space AI at public Instagram profiles in your space for content inspiration.

  3. Content & calendar

    Generate more ideas and edit them in place

    • Ask for more content ideas whenever the calendar runs short
    • Edit an idea's details directly before it becomes a post
  4. Creative

    Better-looking post designs

    • Refreshed templates, a cleaner arrangement of text and images
    • AI illustrations drawn to fit each post and your brand