An AI business advisor for agency owners

§02 / DEFINITION

What this is, in four sentences

A personal AI advisor for owner-operators that cites every claim to a passage you can open, and declines to answer what it cannot cite. You open it and it is a conversation, not a dashboard, and it asks about your agency before it advises you on it. When it can point at a source, the claim carries a chip at the end of the sentence that used it, and the chip opens the passage rather than the document. When it cannot point at anything, you get the search it ran and a logged gap instead of a confident paragraph.

That last sentence is the product. Everything else on this page is detail hanging off it.

§03 / FIT

Who this is for, and who it is not for

This is for one owner running a services business with 0 to 3 staff, who makes the pricing, staffing and offer decisions alone and has roughly twenty minutes at a time to think about them. Agency, studio, consultancy, production company. The shape that matters is one decision maker, not the vertical.

It is not for you if you want an assistant that always has an answer, if you want proof it worked for somebody like you, or if you want a tool your team logs into. There are no customers yet, so there is no proof of the second kind and we are not going to manufacture it. On the third, this is one owner and one screen.

It is also not for you if the decision in front of you is a hire, a firing, a loan, a tax position, a policy or anything with a licensed professional standing behind it. Those are on the list below and it declines all of them.

§04 / THE DIFFERENCE

The one thing that makes it different

It checks whether it can cite an answer before it gives you one. The check runs before the model is called, so on a refusal there is no fluent paragraph waiting behind the decline for you to push at. Here is what that looks like.

That is an illustration of the refusal card, not a captured session. There are no sessions to capture yet. What is real is the behavior in it: the search it names, the three nearest things it holds, and the gap that gets written down.

A general chat subscription answers that same question in four seconds, from nothing, and it sounds right. That is the honest comparison, and it is the whole argument for the price.

There is a second refusal above the first. Eight subjects are rejected before it searches anything at all.

An agency owner asks about hiring constantly, and this refuses every time. That is deliberate. Those are the eight where a confident wrong answer costs more than no answer, and where the person you should be asking holds a license.

§05 / A SESSION

What a Tuesday night with it looks like

Twenty minutes, one question, ending in something written down. Suppose you had already told it that one account takes 42 hours a month and that you target $95 an hour, and you ask whether to raise that retainer from $3,000 to $4,000. The shape of the answer is this.

Those figures are made up, because there are no customer sessions to quote. The mechanics are not. Four things happened there that cannot happen in a general chat window: a named mechanism with a source you can open, your own figures applied to it with their status on the label, a part marked refused instead of filled in, and a decision written down with a date it gets looked at again.

The return date is the part owners underrate. Six weeks later the question is not what did we say. It is did it work, and the record answers it without you reconstructing the conversation from memory.

§06 / MEMORY

What it holds about your agency, and how you correct it

Everything it holds about your business sits on one editable panel. What you sell, who buys it, what you charge, your delivery hours, your margin target. Each fact lands there as you say it, as a chip you can undo on the spot. Edit a line, delete a line, or just say the right number in the chat and correct it there.

Every number carries its status on the line. Self-reported if you said it. From document if it came out of a file you handed over. Stale once it is more than 90 days old, rather than quietly aging into a wrong answer. There is no verified status anywhere in the product, because nothing here has been audited and a label claiming otherwise would be the first lie on the page.

One business or five, each is walled off from the others. Separate facts, separate memory, separate conversations, and nothing from one is ever retrieved into another.

§07 / LIMITS

What it is not

It is not a dashboard. There is nothing to configure before it is useful and no charts to read. Threads on the left, the thread in the middle, a composer at the bottom.

It is not connected to anything. Not your bank, not your CRM, not your ad accounts. It reads no live data and there is no background sync, so nothing updates itself while you sleep. You tell it your numbers or you give it a file, and files today mean markdown, plain text and CSV, read in your browser. PDF, Word and Excel are not supported yet, and we are not going to list them and hope you do not try.

It is also not an agent. It takes no actions on your behalf, holds no accounts, and sends no messages. It advises the person who runs the business. That is the whole scope.

§08 / ALTERNATIVES

The alternatives you are weighing

Three of them, and each has its own page here written to send you away when it is the better buy.

• The chat subscription you already pay for, at roughly a fifth of the price: ai business coach vs chatgpt.
• A human coach, who sells accountability and a calendar rather than a checkable answer: ai business coach vs human coach.
• Nothing at all, which is the right answer more often than this category admits: is an ai business coach worth it.

The recommendation, stated plainly: if you have never paid for advice before, keep the chat subscription and read the worth-it page first. If you have paid for a community or a coach and the problem was that nothing survived the call, this is the one built for that specific failure.

§09 / PRICE

Price and how access works right now

Monthly

$99 / mo

Annual, pre-selected

$79 / mo, billed once at $948 / yr

Trial

7 days, card required

Free tier

none, and there will not be one

Second tier

none

Team seats

none

Status

Pre-launch

Priced in USD and EUR from the first day. EUR is a currency here, not a different product. A free tier would create pressure to answer more, and this is built to answer less.

Accounts are approved by hand and capped while this is small. That is not a scarcity tactic, it is the state of the build. Request access and a person reads the request.

§10 / DATA

Your data and the AI processing

Your data is stored in the EU. AI processing is done by a US-based provider under EU Standard Contractual Clauses, and is never used for training.

What that sentence does not say also matters. The passages and the facts you give it are sent to a US model vendor on every turn. We do not claim zero retention, because it has not been granted to us in writing. We do not claim SOC 2, ISO, end to end encryption, or that your data never leaves, because none of that is true here and you would find out.

This site does not track you. No analytics script, no pixel, no third party font request, no cookie banner, because there is nothing here to consent to.

§11 / SCOPE

What this page does not cover

This page does not reproduce the refusal behavior in full. The eight declined subjects one by one, the second refusal when nothing can be cited, and the refusal card written out completely are on what agi.ceo refuses to answer.

It also does not cover outcomes. There are no hours saved, no revenue added, no accuracy rate and no satisfaction score on this site, because nobody has used the product long enough for any of those to exist. There are zero customers, so there are no logos, no testimonials, no star ratings and no revenue screenshots either. That space stays empty until there is something true to put in it.

§12 / SOURCES

Sources

• Product behavior on this page (the eight declined subjects, the three nearest passages shown on a refusal, the 90 day stale threshold, the absence of a verified status, markdown, text and CSV input only): the agi.ceo build specification, 2026-09-20. These are checkable on day one of a trial.
• Price ($99 a month, $79 a month billed annually at $948 a year, 7 day trial, card required, no free tier): published by us on this site. It is the only price there is.
• Data and processing: our own processing statement, reproduced verbatim above.
• The "roughly a fifth of the price" comparison: a ratio, not a quoted vendor price. We are not publishing a competitor's monthly figure we have not re-checked on the vendor's own pricing page today, on a page whose argument is provenance. Check it yourself before you use it as a number.
• Everything about customers, results and demand: no source, because there is none. Marked Pre-launch throughout.

Checked on 2026-09-20.

§13 / ACCESS

Seven days is enough to find the hole

Open it, tell it what you sell and what you charge, and ask the question you have been circling since last quarter. Either it gives you a mechanism with a source you can open, applied to your own numbers, with a date on the decision. Or it tells you it has nothing it can cite and shows you what it looked for. Both are worth knowing by Friday. Read the whole argument on agi.ceo.

Request access

Approved by hand. Accounts are capped while this is small.