Autopilot · AI performance marketing · software, not a service
It builds the ads. It will not spend on them.
Autopilot researches your market, writes the angles, makes the creative and builds the campaign on Meta. Then it stops, because every ad it makes arrives paused and your budget is yours to release.
We do not run your ads. This is software you operate on your own ad account, with your own budget, and there is no version of it where we hold the card.
- What it is
- Software you run, on your own ad account
- What you set
- The business details once, and the budget
- What it decides alone
- Only what saves money and can be undone
- What it never does
- Start or raise spend without your yes
- Built in
- TypeScript, running as a container
No install count, no spend figure, no return on ad spend anywhere on this page. What is missing, and why, is written out below.
The part that matters
Most tools like this can spend your money. This one cannot.
The rule is not a setting. It is written into what the system is allowed to do at all.
An action either saves money and can be undone, or it does not. Only the first kind runs by itself. Everything else waits, and the switch that allows it to act at all is off until you turn it on, one client at a time.
The action
Who decides
Why
Build a campaign, ad set, creative and ad on Meta
Runs alone
Created paused. It exists, it is reviewable, and it is spending nothing.
Pause an ad set that is losing money
Runs alone
The only kind of action it takes alone: money-saving, reversible, validated against the live campaign, and logged with an undo.
Exclude a placement that is wasting budget
You confirm
You confirm it. The rule will never leave a campaign with zero placements.
Resume anything it paused
Waits for you
Resuming spends money again, so it waits for one tap from you.
Raise a budget or lift a bid cap
Waits for you
Anything that increases spend is an approval, every time, without exception.
Launch an ad so it starts spending
Waits for you
Needs the ad account connected, a budget set by you, and an explicit go.
Which leaves exactly one thing
It can turn spending off by itself. It can never turn spending on.
What a performance marketer does on a Tuesday, done every day.
You give it the business once. It does the rest of this on its own schedule.
16
jobs running today
9 more not working yet
Everything in this list runs now, on the credentials the system already holds. The things that do not run are in their own section further down, named individually and with the reason attached, because a feature list that quietly mixes the two is worth nothing to you.
Research and creative
01
Research
Reads the market before it writes a word
Live search across competitors, the category and the customer’s own language, pulling six to ten verbatim quotes. The sharpest of those becomes the hook, which is why the copy does not read like copy.
02
Angles
Six to eight ad angles, with the copy attached
Each one carries a big idea, a hook, a headline, body copy, a call to action, the awareness stage it is aimed at, and the proof it would need to stand up. Not a list of taglines.
03
Creative
Builds the ad on your own product photo
It ranks the usable photographs on your website and composes headline, subtext, button and logo over the best one, in your brand colour and your brand typeface. A dark logo gets a light chip behind it so it stays legible.
04
Variants
Carousels, awareness variants and A/B sets
Five-slide carousels with or without photography, five variants by awareness stage, or four for a straight A/B. One angle in, a test matrix out.
05
Brand kit
Takes the brand off your URL
Logo, colour palette and typeface read straight from your site, so nothing about your brand has to be typed in twice or guessed at.
06
Language
English, Hinglish or Hindi
Chosen once at setup and applied to every generated word after that: ad copy, captions, video scripts and lead forms.
On the ad account
07
Campaigns
Builds the whole structure, paused
Campaign, ad set, creative and ad. Including a full architecture: one budget across broad, lookalike and retargeting sets with an A/B creative matrix under them. Every piece of it arrives paused.
08
Formats
Lead forms and click-to-buy ads
A native instant form that opens inside Facebook or Instagram with your own qualifying questions on it, or a conversion ad that sends the click to a product page and optimises on purchases where a pixel exists.
09
Segments
Reads age, region, placement, device and hour
The breakdowns a media buyer opens first, read back as which segment is carrying the account and where the budget is leaking.
10
Fatigue
Says how many days an ad has left
Frequency climbing against click-through falling, turned into a forecast rather than a post-mortem, with a fresh creative concept generated on request before the ad dies.
11
Profit
Answers in profit, not in ROAS
Your average order value, margin and close rate become profit, profit-on-ad-spend and a break-even cost per result. A campaign can have a good ROAS and still lose money, and this is the number that shows it.
Deciding, and acting
12
Brain
One ranked list instead of six dashboards
Paid performance, segments, fatigue, profit, organic posts and the site audit are read together, and come back as the situation, today’s focus, the watch-outs and what changed since last time. Every item is tagged with whether it needs you.
13
Hands
It can act, inside the leash above
The per-client switch that lets it act at all is off until you turn it on, and even on, the only thing it does by itself is pause something that is losing money.
Organic, leads and the site
14
Organic
Daily posts, written per platform
Separate captions for Facebook, Instagram and LinkedIn rather than one caption pasted three times, with a branded image and rotating content pillars. Either it posts them, or it queues them for you to approve.
15
Website
Audits the site the ads point at
Core Web Vitals, then a crawl of the home page, a product page and cart or contact, then a scored conversion report that names why people are not converting and writes out the fixes. Ads are wasted on a broken page.
16
Leads
Scores each lead and hands it on
Every lead scored nought to a hundred and tagged hot, warm or cold, with an optional webhook that pushes it into a CRM.
Written down rather than left out
Nine things this does not do yet.
Code being finished is not the same as a feature working, and we are not going to blur the two.
For most of these the code is written and reviewed and waiting on an access token, an approval or an API key that has not been granted. One of them is genuinely not finished and is marked as such. If a demo you are shown elsewhere includes any of these, ask which credential is behind it.
01
Your store’s real revenue, order value and repeat rate
The Shopify connection is written and the profit maths is waiting for it. It needs an Admin API token, and no token has been issued. Until it is, profit is computed from figures you enter rather than from your orders.
02
Google Analytics 4
Sessions, bounce and which country, device and channel actually convert. Needs a property ID and an OAuth grant that do not exist yet.
03
Google Ads
Connecting an account and reading its campaigns is built. Google has to approve a developer token before any of it runs, and creating campaigns there is a further step after that.
04
Watching a competitor’s live ads
Pulling a competitor’s running ads, spotting the ones that have run long enough to be winning and alerting on them, all needs separate Ad Library authorisation. Coverage also varies by country, so this may stay partial even once it is granted.
05
WhatsApp, in both directions
The morning push, and running the system by replying to a message, both need WhatsApp Cloud API credentials and a webhook subscription. Neither is in place.
06
Posting to a LinkedIn company page
Posting to a personal profile works today. A company page needs LinkedIn’s Community Management API, and that approval has not been granted. Until it is, the company page is copy and paste.
07
Replying to comments on Facebook and Instagram
The code reads a comment in context, drafts a reply and holds anything about a complaint, a price or a serious lead for a human. It needs three permissions on the page token that are not on it. On LinkedIn, reading and replying to comments is not possible at all.
08
Video ads
The furthest from ready and the most honest thing on this page. Video from the real product, a garment on an AI model, and a synthetic presenter reading a script are all written, and not one of them has been confirmed working end to end on a single finished video. It is parked, not shipped.
09
Signing up for it yourself
There are no self-serve accounts, no one-click connections and no billing. Setup happens with us in the room. That is deliberate and it is last on the list, but it means this is not a link you can go and buy today.
That list will get shorter. It will be edited here when it does, rather than quietly becoming true somewhere else.
The number it is built to
It was designed against a price, not priced after it was designed.
₹12,000 a month
our target price · not a published price list · no billing in the product yet
The comparison in our heads was a performance marketer at forty to fifty thousand rupees a month.
That second figure is our own assumption about what the role costs, taken from our commercial judgement rather than from a survey, and we are not going to dress it up as research. Price your own alternative before you accept it from us. It is a constraint we set for ourselves at the start, which is why the system leans on free reads and on doing the expensive work once rather than on every refresh.
It is also not a proven saving. Nobody has run this against a marketer’s salary for a year and measured the difference, so treat the twelve thousand as what we are building to, and the rest as intent rather than as a result.
What this page does not claim.
There is no user count here, no budget managed, no cost per result and no return on ad spend.
Nobody independent has audited a number from this system, and any figure we picked would be one we also got to choose the account for. There are no client names on this page either, and no logos. The businesses it has been pointed at did not agree to appear in our marketing, and being able to name them is not the same as being allowed to.
What is here instead is what it does, what it refuses to do, and what it cannot do yet. The other software we have built and run is presented to the same standard, and the AI agents we build for other people’s operations are held to the rule that produced the guardrail above: an agent does not get to decide it is finished.
When there is a figure with an instrument behind it, it will appear here with the date it was read, the way the two case studies we can evidence already do.
The questions that decide whether this is for you.
Do you run my ads for me?
No. Autopilot is software you run on your own ad account, with your own budget, and we do not run anyone’s advertising as a service. That line is the same everywhere on this site: we build and automate stores, and we do not buy media. If what you want is a person to hand your budget to, we are the wrong supplier and you should hire a media buyer.
Can it spend my money without asking?
It cannot start spend or increase it. Ads are built paused, so nothing goes live until you say so. Once the per-client switch is on, the only action it takes on its own is pausing something that is losing money, which is reversible and logged with an undo. Resuming, raising a budget, lifting a bid cap and launching an ad all wait for a tap from you. The switch is off by default.
What does it cost?
The number it is being built to is twelve thousand rupees a month. That is a target we set ourselves rather than a published price list, because the setup is still done with you by hand and there is no billing in the product yet. If you want a figure to plan against, use that one and tell us if the scope you have in mind breaks it.
Which parts genuinely work today?
Research, angles and copy, static creative on your own product photography, carousels and variants, the whole Meta campaign structure built paused, segment and placement reads, fatigue forecasting, the profit view, the decision layer and its guarded actions, daily organic posts to Facebook, Instagram and a LinkedIn profile, the website conversion audit, and lead scoring. The list of what does not work yet is on this page too, with the reason for each.
Why is there no case study or performance figure here?
Because we do not have an audited one. Every number you would want here, spend managed, cost per result, return on ad spend, is a number nobody independent has checked, and one we could shape by choosing which account to quote. The rest of our software is presented the same way on the apps page. When there is a figure with an instrument behind it, it will appear with the date it was read.
Is it a Shopify app?
No. It runs against Meta, against your website and against your own business numbers, and it has no Shopify surface. It is listed with our apps because it is software we designed, built and run, which is what that page is a record of. The Shopify connection described as missing above would read your store’s numbers into it, and that is a data feed rather than a Shopify app.
Who is it actually for?
A business already spending on Meta, or ready to, that wants the work of a performance marketer done consistently rather than the judgement of one replaced. It suits an owner who will look at a short list of decisions each morning and tap yes or no. It does not suit anyone who wants to never look at their ad account again, because the guardrail that makes it safe is precisely the thing that needs you.
Tell us what you are spending on Meta, and what you wish somebody watched.
We will tell you which parts of this would apply to you and which of the missing pieces you would actually feel.
No demo environment built to look good, and no login you can sign up for today, because there is not one to give you.
What we ask you first
01
What you would be running it on
02
Which of the missing pieces you actually need
03
What your ad account looks like today
04
Whether the guardrail suits how you work
Autopilot sits alongside ProfitOS and LineDrop on the apps we build and run. If the job you have is a repeating operation rather than advertising, the automations we already ship are the cheaper answer.