An AI audit week, day by day
A one-week AI audit, day by day. The calls, the three hours it asks, the report you own.
“AI audit” sounds vague, and plenty of them are. The useful kind has a timetable, a fixed cost in your time, and a deliverable you own. Here is what a one-week audit involves, day by day.
This walkthrough describes the audit Opafex runs, because that is the one we can describe honestly. But the structure is worth knowing even if you never buy one. It tells you what to expect from any provider, and what to push back on.
What an audit is not
Before the timetable, the exclusions. A useful AI audit is not a strategy deck, a maturity score, a vendor shortlist, or a sales pitch for follow-on work. The category is full of all four. Each one tells you something about AI in general and nothing about your business in particular.
The question an audit should answer is narrower and harder: which work should a machine do, and which should it not? Everything in the week below serves that question.
The week, day by day
- —Day 0 — intake, 15 minutes. A short form: facts about the business, plus a handful of real working documents. Real documents matter. People describe a process as it should be; documents show what it is.
- —Day 1 is the owner call. It runs 90 minutes. A call about the whole business, with a screen-share of how work moves. Where orders arrive, what gets retyped, what waits on one person.
- —Day 2 is the operations call. It runs 30 minutes. A conversation with someone who does the day-to-day work. This call produces the best findings. The owner knows the intended process. The operator knows the workarounds, the spreadsheet on the side, and what they already paste into ChatGPT.
- —Days 3 to 5 — analysis and drafting. Nothing needed from you beyond the occasional short email question as gaps appear.
- —The report arrives on day 7. We walk you through it in 45 minutes. The written report arrives and gets explained on a call, line by line where you want it.
Total client time: three hours, across four sittings. That cap is a design constraint, not an estimate. No data extracts, no system access, no preparation.
Why the time cap matters
Engagements die of homework. An audit that asks your team for data extracts and pre-reading competes with the day job, and the day job wins. If a provider needs more than a few hours of your time, ask what the extra hours buy.
The core output: every workflow, allocated
The centre of the report is an allocation map. Every recurring workflow in the business gets placed in one of four columns:
- —Machine work — runs unattended. High volume, low judgement, easy to check.
- —Hybrid — a machine drafts, a person signs off. Most of the interesting work lands here.
- —Human work — automating it would cost more than it saves, or the judgement is the product.
- —Leave alone — not worth touching now, and possibly not ever. Naming this column out loud is half the value.
The placement matters less than the reasoning. Each line should carry the argument for its column, so you can weigh every call yourself and overrule any of them. An allocation with no reasoning shown is an opinion, not an audit.
What the report contains
Five sections, in this order:
- —Where the time goes — the recurring work across the business, quantified, in your own people’s words.
- —The allocation map — every workflow placed, with reasoning per line.
- —The top three moves — written out as actions, each with honest effort and cost estimates attached.
- —Quick wins you can do yourself — steps to take this month without hiring anyone.
- —A data and governance note — where your data would go for each opportunity, and what that rules out.
You own the report. That should be explicit before you book. A deliverable you cannot show to another supplier is a brochure.
What happens after
Most audits do not end with the auditor building anything. The report may recommend an off-the-shelf tool, a change to a process, or hiring a person. Those answers turn up more often than a build does, and a good audit says so up front. If you want the fuller decision framework, see how to decide whether to build, buy, or change the process.
The cost, and who carries the risk
The Opafex audit is £2,500, fixed, paid 50% on booking and 50% on delivery. Fixed matters more than the number. A day rate makes a harder audit your invoice; a fixed price makes it the auditor’s problem. For the wider picture on pricing models, see what AI consultancy costs in the UK.
And if you want the detail on the audit itself: the audit page lays out the whole week, the coverage, and the fit test.
Frequently asked questions
How long is an AI audit?
A focused audit of one business takes a week. Intake starts it, two calls follow, then analysis and a written report on day 7. Longer engagements exist, but a week maps the recurring work of a small or mid-sized business.
How much of my time?
A well-designed audit asks for around three hours, spread across four sittings. An intake form, an owner call, an operator call and a report walkthrough. It should not ask for data extracts, system access or homework.
What do you get?
A written report you own, with the reasoning shown. It quantifies where time goes, allocates every recurring workflow and notes where your data would go. It costs the top recommendations and lists quick wins you can do yourself.
Is it a sales pitch?
It should not be, and you can check before booking. Ask whether most audits end with the auditor building something. Often the right answer is an off-the-shelf tool, a process change or a hire.
What does an AI audit cost?
It varies by provider and pricing model. The Opafex audit is £2,500 fixed, 50% on booking and 50% on delivery. An audit that turns out harder is the auditor's problem.
Keep reading
Get started
See it on your own business
Opafex is live in beta. Tell us your first agent’s job and we’ll set you up.