AI guidance for small business: the diagnosis comes first.
Machines have taken over the busywork of analysis. That isn't the point. The point is what the freed-up human time gets spent on, and we spend it sitting with your business long enough to find what is actually wrong: not the problem that is easiest to invoice, not the problem your team has already decided it has. You leave with a written proposal that names the real problem and argues for it properly. AI comes after that, if it comes in at all.
Most of this market sells the treatment and works backwards to a diagnosis. We run it the other way round.
- You bring the situation: what your business does, where the hours go, and anything a vendor has already put in front of you. The machines chew through the busywork of the analysis, the numbers, the comparisons, the reading, so none of your fee is spent on that.
- The human hours go into understanding your operation. We sit with it long enough to get past the obvious candidates and find the problem nobody has dared look at, because looking takes time nobody had.
- You get a written proposal that names the real problem and argues for it, with the justification laid out so you can check the reasoning yourself. Only then does AI enter, as the means of tackling it, and only if the case for it holds.
A proposal can reach four kinds of verdict.
The most common outcome: the problem worth fixing is named and argued for, and one or two AI moves genuinely fit it, each with a real cost, who runs it, and the condition that makes it a win.
Sometimes the diagnosis holds but the treatment is a process change, a hire, or a hard conversation. The proposal says so and argues for that instead. We don't bend the answer towards the tools.
The thing a vendor pitched, or the thing the team has agreed to believe, doesn't survive the looking. The proposal says what does, with the reasoning laid out so you can close those conversations cleanly.
A live, respectable verdict, not a failure of the engagement. If the honest reading is that your operation has nothing AI would meaningfully improve right now, the proposal says that in writing, and you keep it for the day that changes.
One fee, published here, no surprises.
A single fixed £950, whatever the verdict turns out to be. What the fee buys is human attention; the machines cover the donkey work. Most owners weighing this are looking at vendor proposals that start in the low thousands and were written to justify a build. £950 for a diagnosis from someone with no stake in the treatment is a small fraction of that. The price sits on this page in plain sight, because a firm that sells you an honest diagnosis shouldn't make you email to find out what it charges.
Leave with a written proposal that names the problem worth fixing and argues for it, even if the verdict is that you don't need AI yet, or you do not pay.
We think guidance is only worth having when the giver has nothing to gain from any particular answer. So:
- There is no build waiting at the end to upsell you. If a done-for-you job genuinely fits later, we will say so, but the guidance does not lean that way to set it up.
- We don't resell anyone's software, so no vendor pays us to point you at their tool.
- No commission, no referral fee, no kickback from anyone you end up working with.
- The £950 is the only money that changes hands on this. If the honest answer is to do very little, we have still been paid, and so we can afford to mean it.
Two examples from our own desk rather than a brochure. A side venture we liked went through the full diagnosis before launch: the unit economics failed the numbers test and it was killed before a pound of ad money was spent. A separate run stress-tested a five-figure acquisition plan and redirected the budget after the claims behind it did not survive checking. And when dispute paperwork had to go out, every factual claim was refute-checked against filed documents first: twelve of fourteen new claims confirmed, two discarded before use. That is the standard the proposal is held to: every load-bearing claim marked supported, refuted or unverifiable before you act on it.
A written proposal, and the sitting-with-you that produces it. The proposal is the artefact you keep: it names the problem actually worth fixing, lays out the justification in full, and specifies the treatment, AI or otherwise. It is written so that any competent implementer could act on it, not just us. Readable by you, forwardable to a co-owner or your accountant without translation. No slide deck, no jargon, no upsell at the bottom.
The session is booked at your convenience once we have spoken. The written plan usually lands within five working days of the session. If your situation is bigger than a single call, or a decision deadline genuinely will not wait, say so when you book and we will be honest about what fits.
Tell us what you think is wrong.
Email hi@cooltake.ai with a line on your business and what you think the problem is. Part of the job is testing whether that holds. We'll confirm the £950 and how it runs, then do the looking and write the proposal. Or start with a short call, twenty minutes, no obligation.