What an AI readiness audit is, what it should contain, and what UK firms charge for one.

An AI readiness audit, also sold as an AI adoption audit, opportunity audit or readiness assessment, is a short fixed-scope engagement that answers three questions: where does time and money go in this business, which of that could AI take on, and what would each candidate cost to build and to run. It ends in a written document you keep whether or not you buy anything else. Done well it is the cheapest thing in the AI market, because it stops the expensive things: the pilot with no baseline, the tool bought for a problem nobody measured, the build that costs more to run than it saves.

It is not a sales call with a report attached, though several free ones are exactly that. It is not a technology audit of your network, though the readiness of your systems is one of the things it scores. And it is not a strategy workshop: a good audit produces a ranked list with prices, not a vision, and the test of one is whether a different supplier could build from it without asking a question.

The market for these in the UK has settled into a recognisable shape, with published prices from free to £8,000 and a fairly consistent list of contents. This guide sets out the contents an audit should have, the published prices side by side, a table you can score your own business on before paying anyone, and a worked example of what the output looks like.

The names differ from firm to firm; the contents should not. If a proposal lacks more than two of these, it is a lighter product than the word audit implies, and should be priced as one. The list is drawn from what the paid UK providers in the next section publish as their deliverables, reconciled, with the two items most of them omit, the measured baseline and the running cost, added because they are the two the build will stand or fall on.

These are the firms whose audit prices we could read on their own pages on 14 September 2026, excluding VAT. Where a firm publishes no price, the table says so. The free products are included because they are what most searches for the term return, and because it is worth knowing what they are. The final column, whether the fee is credited against a later build, is the one buyers most often forget to ask about and the one that most changes the real price.

All figures exclude VAT and were read on the providers’ own pages on the date shown. The two free products are a score and a calculator respectively, not an audit in the sense of the ten deliverables above.

A £495 product and a £5,000 product are both called an audit, and the difference is days. Reading the published contents, the cheaper products are one or two days of a consultant’s time and a template: a questionnaire, a score, a short report. The £2,000 to £5,000 products are five to fifteen days: interviews, a look at real data, a ranked list with numbers. Set against the measured contract market, where the median AI consultant contract day rate in the six months to 14 September 2026 was £575 according to IT Jobs Watch, a £2,450 audit is about four days and a £5,000 audit about nine. That is the sanity check to apply to any quote: how many days does this imply, and could the ten deliverables be produced in them?

The other variable is what happens afterwards. Wingenious credits half its fee against work booked within sixty days; we credit all of ours against any build; most publish nothing. A credit is not a discount, it is an alignment: a firm that will earn the fee back only if it recommends something you go on to build has an incentive to recommend something, which is why the document should include the "do nothing" option and why you should read it as a buyer, not a client.

Free assessments have their place. Techcare’s produces a score in minutes and is honest that it is oriented to Copilot adoption; Digital Eight’s is a missed-calls calculator that leads to a demo. Neither will tell you which of your processes to automate first, and neither times anything, but either will tell you in ten minutes whether you are ready to have that conversation, which is worth knowing before spending £2,500.

Before paying for an audit, score your business on the eight dimensions below, one, three or five each. The total tells you what to buy. Under sixteen: fix the basics, most of which cost nothing, and take a free assessment. Sixteen to twenty-eight: a paid audit will find things, and its recommendations can be acted on. Over twenty-eight: you probably already know the first thing to build, and an audit that doubles as the scoping of that build is the better purchase.

Our own scoring, not a published standard. DSIT finds 54% of current UK adopters feel ready to scale their use and only 34% of planned adopters feel ready to implement; the table is a way of finding out which you are before anyone charges you to tell you.

DSIT’s AI Adoption Research, 3,500 UK businesses interviewed between February and May 2025 and published in February 2026, is the best map of why UK firms do not adopt. Among the factors preventing adoption, 71 per cent cited a lack of identified need, 60 per cent limited AI skills, and 48 per cent a lack of tools or platforms. Asked how significant particular barriers were, 80 per cent called ethical concerns significant, 76 per cent high costs, and 72 per cent unclear regulation. Nearly half, 47 per cent, of businesses planning to adopt had not budgeted a specific amount.

The first two are what an audit is for. "No identified need" is rarely true of a business that keys data between systems by hand; it is usually a measurement gap, and the baseline is the deliverable that closes it. The skills barrier is real, and the ONS finds the sensible response is common: of businesses citing a lack of expertise, around 62 per cent are training or retraining existing staff. An audit should say which skills you need in-house after handover and which you can buy by the project.

The ONS series puts the size of the gap in one line. In June 2026, 35 per cent of UK businesses with ten or more staff used AI, but only one adopter in ten used it extensively. Among the smallest firms it was 28 per cent. Most UK adoption is one person with a chatbot. The audit is the step between that and a system that runs a process, and DSIT’s finding that 65 per cent of firms plan to adopt off-the-shelf applications suggests the honest recommendation will often be to buy rather than build.

An illustration built from published figures and stated assumptions, not a client account. A 35-person building-services contractor asks what AI could do for it. The audit interviews the office team, times three tasks for a week, and reads the systems. The output, condensed, looks like this; the full version is a document of perhaps fifteen pages, and the part that takes longest to produce is the first line, because nobody had timed anything before.

If you have fewer than ten staff and no process that eats hours in a predictable shape, an audit will not pay back. Score yourself on the table, buy business accounts for the tools your team already uses, write the one-page policy, and see what a month of that produces. The ONS finds cost is cited as a barrier by only 7 to 14 per cent of businesses; the commoner problem is not knowing what to do first, and at that size the answer is usually visible without help.

Equally, if you already know the process, have the baseline and have a budget, an audit that ends in a report you then have to re-explain to a builder is a delay. Buy the scoping of the build instead, from a firm that will build it, on a fixed price, with the audit credited. That is the reason our audit is credited in full: a client who already knows the answer should not pay twice for it.

For transparency, since we are a row in the table: a fixed-price AI adoption audit at £2,500 to £6,000 over one to two weeks, containing the ten deliverables above, ending in a written report on how the operation actually runs, opportunities ranked by return each with an indicative price and timeline, an ROI model with running costs included, and a prioritised backlog you keep whether or not you build with us. The fee is credited in full against any build that follows: a first implementation at £4,000 to £15,000, a custom AI system at £8,000 to £25,000, or bespoke software from £8,000. No day rates, no minimum, and the report says "do nothing" when that is the honest answer.

We read the pricing and service pages of every UK provider we could find that publishes an audit price, on 14 September 2026, and recorded each figure with what the page says it includes. Where a firm quotes only "tailored" pricing we recorded that. All prices exclude VAT. Adoption and barrier statistics are DSIT’s AI Adoption Research, published 13 February 2026 with fieldwork from February to May 2025, and the ONS article on AI in UK businesses published 20 July 2026. The self-assessment table is our own and is not a published standard.

The worked example uses stated assumptions, our published price ladder, and model and messaging prices read on the OpenAI, Anthropic and Twilio pricing pages on the same day; it is not a client account, and the hourly cost in it should be replaced with your own. We will revise the table as the sources change; the date at the top of the page is the date of the last check.