The AI Brain Audit: what actually happens in a half-day diagnostic

The AI Brain Audit started as something I did for free, then started doing for £200, and now sits at £700. The price went up because I kept underestimating how much I was actually shipping. A half-day diagnostic, a written report, a prioritised 90-day roadmap, and a 30-minute follow-up call. That’s a real deliverable, not a sales conversation in a different jacket.

This post is the version of the Audit pitch I wish I’d written six months ago. If you’re sitting on the AI Brain page wondering whether £700 is worth a half-day of someone’s time, this should tell you. And if you read it and decide the Audit isn’t the right move yet, I’ll point you at the right one.

Why an Audit rather than just a chat

The honest answer: because chats don’t ship anything. I have plenty of free 30-minute calls. They’re useful. You leave with a clearer picture of your situation and, often, a recommendation you can act on without me. That’s how about half of them end, and I’m fine with that.

What a 30-minute call cannot do is look at your actual setup. Your Claude projects, the way your team is using them, the prompts you’ve copy-pasted into a Notion doc, the workflows that exist only in one founder’s head. By the end of a 30-minute call I have a hypothesis. By the end of a half-day Audit I have a diagnosis.

That distinction matters because the next move you make costs real money. A £6,500 Sprint is the right answer for some founders. A £2,500 Starter is right for others. And for a meaningful number of people, the right answer is “cancel two of your three subscriptions, set up one project properly, and read the prompt library.” The Audit is the cheapest way to find out which of those three you are.

Who the Audit is for (and who it isn’t)

The Audit is built for founders and small teams who have already tried. You’re using Claude or ChatGPT (or both) every day. You have a sense that the output isn’t compounding, that the gap between your setup and “AI-native” is wider than it should be, and you don’t trust the LinkedIn pitches from generic AI consultants. You want someone who has built and shipped to look at what you’ve got and tell you where the leverage actually is.

It is not for founders who haven’t started yet. If you don’t have an AI subscription, the right next move is the 5% post and a £20/month Claude plan. Use it for two weeks. Then come back. There’s no diagnostic worth running on an empty stack.

It is also not for engineering teams looking for technical implementation. The Audit is a founder and ops-level conversation. If you need a coding agent strategy, an MCP server design, or an eval harness, that’s a different scope and a different conversation. Tell me on the pre-call and I’ll route you.

Step 1: The 30-minute pre-call

The half-day is paid. The pre-call isn’t. We do it a week before so the half-day doesn’t get spent on context I could have collected by email.

What we cover, briefly:

• Who you are. Stage, team size, what you actually do all day. Two minutes.
• What you’ve tried. The tools you pay for, the workflows you’ve attempted, the one or two that stuck and the four or five that didn’t.
• What “good” looks like. Concretely. “I want to stop spending Sunday evenings on investor updates” is a better answer than “I want to be more productive.”
• The honest constraint. Budget, time, who actually has to use this thing afterwards, whether your co-founder is on board or sceptical.

I send a short prep doc back the same day. It lists the five things I want to look at during the Audit, asks for read access to the projects or workspaces I’ll need, and confirms the date and time. If you decide after the pre-call that the Audit isn’t the right move, you walk away without paying anything. I’d rather lose the £700 than do an Audit that wasn’t going to land.

Step 2: The half-day itself

Four hours, on a single call, on the day we agreed. No multi-week scheduling, no follow-up emails to chase missing inputs. The point of a half-day is that it’s a half-day.

Here’s the agenda I run, give or take half an hour either way depending on what we find:

Hour 1: The current state, in front of me

We screen-share through your actual setup. Your Claude projects (or ChatGPT projects, or both). The system prompts. The knowledge files. The connectors that are wired up and the ones that should be but aren’t. The workflow you reach for most often. The one you wish worked but doesn’t.

I’m not auditing you. I’m auditing the system. The pattern I see most often is a project with a 4,000-word system prompt, no knowledge files, three workflows attempted and abandoned, and one workflow that actually works but is buried at the bottom of a chat history nobody can find. That’s not a failure of effort. It’s the predictable shape of a system built one Sunday evening at a time.

Hour 2: The workflows, mapped honestly

We list every recurring task in your week that involves you typing into a chat window. Investor updates, content drafts, customer follow-ups, hiring rubrics, all of it. For each one we tag three things: how often you do it, how long it takes today, and how much of the output you rewrite by hand.

The rewrite tax is the one most people don’t see until it’s on a page in front of them. If you’re rewriting 60% of every AI output, the time saving isn’t 90%. It’s closer to 40%. That maths matters because the workflows worth automating first are the ones where the rewrite tax is already low, or where it can be driven low with a sharper persona file.

Hour 3: The voice and knowledge audit

Two artefacts decide whether your AI Brain sounds like you or sounds like the internet. Your style file (how you write, what you’d never say, the words that make you wince) and your knowledge base (what the model can reference about you, your customers, your product). Most founders have neither. Some have one. Almost nobody has both, tuned.

We work through what each one should contain for your situation, and where the source material already exists. Usually it’s there: in old investor updates, in Slack threads, in a Notion doc nobody’s looked at since the founding team meeting. The work is in the extraction, not the writing.

Hour 4: The roadmap conversation

The last hour is the most useful one. By now I have enough to have an opinion. I share the three to five highest-leverage moves, in order, and we talk through each one. What it costs in your time. What it costs in mine if you want me to build it. What you can do this week, this month, this quarter.

The order matters more than the list. Most people pick four things to do at once and finish none of them. The roadmap is sequenced so the first move makes the second move easier, and so on. By the end of the call you have a 90-day plan you could in principle run without me.

Step 3: The written report

You get the report within five working days. About six to ten pages, written for you to share with a co-founder or an exec team if you want. The structure is consistent across every Audit:

1. Diagnosis. One page. What’s working, what isn’t, the one or two patterns that explain most of it.
2. The workflow inventory. The table we built together in hour two, with my notes on which ones are worth automating, which are worth keeping manual, and which should be killed.
3. The voice and knowledge gap. What your style file and knowledge base should contain, with a starter outline and two or three example entries drafted in your voice.
4. The 90-day roadmap. Three to five moves, sequenced, with the why for each one and a rough estimate of effort.
5. The build-or-buy call. An honest take on whether you should run the roadmap yourself, hire it out (not necessarily to me), or do a mix. If any of the moves are better handled by a tool you’re already paying for, I say so.

Two weeks after I send the report, we do a 30-minute follow-up call. That’s there to answer the questions you didn’t know to ask on the day, and to course-correct if you’ve started running the roadmap and hit something. The follow-up is included. I don’t sell it as an add-on.

Step 4: The 60-day credit window

If the Audit makes the case that a Starter or Sprint is the right next move and you book it within 60 days of the report, the £700 comes off the price. That’s not a sales pressure tactic. It exists because some Audits genuinely conclude “you should hire me to build this with you” and I don’t want to charge you twice for the same diagnostic.

If the Audit concludes you should run the roadmap yourself, that’s a successful Audit and there’s nothing to credit. About a third of Audits end this way. I’d rather have a happy founder who didn’t need me again than a confused one paying for a Sprint they couldn’t use.

When the Audit is the wrong thing to buy

Three situations.

You haven’t started. If you don’t have a Claude or ChatGPT subscription, or you’ve had one for less than a fortnight and barely touched it, an Audit will tell you things you’d have figured out for free with two weeks of use. Spend the £20, read the foundation posts on this blog, then come back if you still want one.

You already know what you want to build. If you’ve done the diagnostic yourself and you’ve got a clear list of three to five things you want installed, you don’t need an Audit. You need a Starter or Sprint. We can scope it on a 30-minute call.

You’re hoping to outsource the decision. The Audit makes you smarter about your own setup. It doesn’t take the decision off your plate. If what you actually want is someone to come in and do the AI thing on your behalf without you being involved, the Audit will frustrate you. So will every other engagement I run. Honest signal: this is probably not the practice you want.

If this fits, here’s how to start

The booking flow is deliberately short. You fill in the form on the AI Brain page (one screen, four fields), I send back two or three pre-call slots that week, and we go from there. The half-day usually lands within two to three weeks of the pre-call, depending on diaries.

If you’d rather talk it through first, the free 30-minute discovery call is at cal.com/nick-tong-cto/30min. Pick a slot, tell me a sentence or two about the situation, and we’ll work out together whether the Audit is the right move or whether you should skip it and do something cheaper or bigger.

If you’ve already read the 5% post and the agent stack post and you know you want a diagnostic, the Audit tier is the one to pick on the AI Brain page.