What a Knowledge and Intelligence Audit Actually Involves, and What It Costs

The question that ends most first conversations about AI is some version of "so what would you actually do first?". It is the right question, and the honest answer is usually that nobody in the room knows yet, including me. That is what a knowledge and intelligence audit is for.

This post is a plain description of what that piece of work involves, what you hold at the end, how long it takes and what it costs in pounds. It also covers when you should not buy one, because that turns out to matter as much.

What it is, and what it is not

A knowledge and intelligence audit is a short diagnostic. It establishes where your hours actually go, which of your repeated decisions are written down, and which of them live only in somebody's head. It ends with a prioritised view of what to build first and what to leave alone.

It is not a strategy document, and it is not a technology review. Nobody needs another deck describing the potential of AI. It is closer to what a surveyor does before a building job: someone spends a short, concentrated period looking at how things really work, and writes down what they find, including the parts you would rather they had not.

The distinguishing feature of a good one is narrowness. It looks at one or two processes properly rather than every function superficially, because a map of one real process is actionable and a survey of eleven is a filing exercise.

Why the diagnosis is the bottleneck

There is a common assumption that businesses stall on AI because the technology is hard or the budget is not there. The evidence points somewhere less glamorous.

The Office for National Statistics looked at management practices and technology adoption across UK firms. The most common barriers to AI adoption reported by firms in 2023 were difficulty identifying activities or business use cases (39%), cost (21%) and level of AI expertise and skills (16%). Working out what to point it at is nearly twice the barrier that cost is, and getting on for three times the barrier that expertise is.

That is a diagnosis problem, not a technology problem, and it is exactly what an audit resolves. It sits alongside the wider finding that businesses frequently cannot identify a need for AI in the first place, which is the same gap seen from further back.

The follow-on question is why businesses do not simply work it out themselves, and the government's own AI adoption research answers it without much ceremony. Some businesses highlighted a lack of available time to look into potential AI solutions, explaining that staff are busy with day-to-day tasks, leaving no time to research or explore AI options. The people who understand the process well enough to map it are the same people doing the work. That is the whole reason this gets bought rather than done in-house.

What actually gets looked at

The government's adoption plan for professional and business services puts the readiness gap in numbers. Three-quarters are not yet ready on core enablers such as data, orchestration, and monitoring, and 70% report limited progress on process redesign, even as staff adopt AI tools on their own initiative. Its verdict on the shape of that gap is worth quoting directly. This mismatch creates a sequencing risk: with workforce capability outpacing system and process readiness.

An audit is the thing that fixes the sequence. In practice it covers four areas.

Where the hours go. One or two jobs traced from the moment they arrive to the moment they are done, including every inbox they pass through and every point where somebody waits. The finding is almost never where the leadership team expects, and the person who owns the process is usually the most surprised.

What is written down and what is not. Which decisions exist as reasoning somewhere findable, and which are reconstructed from memory each time. This is where key person exposure usually surfaces, because it becomes obvious how much depends on one person being available.

Where your knowledge currently lives. Which system holds which slice, and what comes back out of each in a usable form. Most leadership teams have never seen this laid out, and the inventory itself tends to be the argument.

What has already been tried. Any previous pilot that stalled, and specifically why. This is the cheapest finding in the whole exercise and the one most often skipped.

What comes out of it

You should end up holding three things, all of them yours to keep and hand to anybody, including a supplier who is not us.

The first is a written map of the processes examined, with the real timings on it. The second is a prioritised shortlist of candidate projects, each with a rough scope and an estimate of the hours it would free, and the third is an inventory of the undocumented reasoning, with a view on what to write down first.

A worked example of why the map matters more than the recommendation: when we mapped the lead process for a global biometrics business, the constraint was not what anyone assumed. Research took over 30 minutes per lead and came down to about 90 seconds, and the bottleneck was never the reasoning speed but the human cost of switching between five different tabs to assemble a single brief. Nobody would have specified that from a strategy conversation. It came out of watching the work.

If you want a rough figure before speaking to anyone, the capacity calculator does the hours arithmetic in a couple of minutes and costs nothing.

Three shapes, and what each costs

Shape	What it covers	Elapsed time	Typical cost
Audit only	One or two processes mapped, prioritised shortlist, knowledge inventory	1 to 2 weeks	From around £3,000
Audit plus roadmap	The above, plus a phased plan with sequencing, owners and success measures	2 to 3 weeks	From around £3,000, scaling with scope
Audit plus first build	The above, plus one automation or intelligence system delivered and running	6 to 10 weeks	£6,000 to £25,000 for the first phase

Two notes on those figures. The spread on a first build depends mostly on how many systems have to be connected and how much of your reasoning is already written down, which is precisely what the audit establishes, so the estimate gets more accurate after it rather than before. And we do not take engagements much below the £3,000 mark, because the time cost of discovery and onboarding makes them uneconomic for both sides.

Judge any of it against the hours it frees rather than as a software purchase. If a proposal cannot show you where the time comes back, the scope is wrong rather than the price.

When you should not buy one

This is the part most agencies leave out, so let me be direct about it.

If you already know which task is eating the hours, and that task sits in one place with a stable process behind it, do not buy an audit. Scope the build instead. The five plain readiness questions in our capability-plan post will get you most of the way for nothing, and commissioning a large diagnostic when you already have the answer is the thing that gives this category a bad name.

The modest end is real. One angel investment community we worked with needed a single human-in-the-loop email workflow, built on the tools the team already used, which gave them four hours back every week. That did not need a discovery programme in front of it. It needed somebody to build the obvious thing.

An audit earns its money in the opposite case: several candidate problems and no agreement on which is worst, work crossing three departments so nobody can see the whole path, or a previous attempt that stalled where nobody can explain why. Paying to find out where the problem is only makes sense when you genuinely do not know.

There is decent evidence that the businesses which do this deliberately are the ones that follow through. The same ONS analysis makes the point directly. Firms with higher management practice scores were more likely to follow through in adopting AI in 2024, given that they planned to adopt AI in 2023. Intent is common. Structured follow-through is what separates the firms that get something running from the firms still discussing it a year later.

Practical takeaways

1. Buy a diagnosis, not a strategy. The deliverable should be a map of how work actually flows, with timings. Recommendations without that underneath them are a sales document.
2. Insist the output is portable. Everything should be yours in a readable format, usable by a different supplier. If it only makes sense inside one agency's process, it is not an asset.
3. Keep it narrow. One or two processes properly beats eleven superficially. A wide survey ages out before anybody acts on it.
4. Check the price against the tier. A focused audit starts from around £3,000 and a first build phase runs £6,000 to £25,000. Six figures to assess readiness is a different product.
5. Skip it when you already know. If the bottleneck is obvious and contained, scope the build and keep the money.

The point of the exercise is not the document. It is that the argument about what to do first stops being a matter of opinion, and once that is settled, the order in which things get written down and built becomes a much simpler conversation. If you would rather work through the trade-offs of doing this with your own team, that comparison lands differently depending on what your team already does, and our AI automation services page sets out how we run engagements.

Frequently asked questions

What is a knowledge audit?

It is a short, paid piece of work that establishes where your business actually spends its hours, which of its repeated decisions are written down, and which sit only in people's heads. It is a diagnosis rather than a build. The output is a map of how work really flows through your business, a shortlist of the places where an automation or an intelligence system would pay for itself, and an honest note on the places where it would not. It is deliberately narrow. An audit that tries to cover every function at once produces a document nobody acts on, which is the common failure mode of the six-figure version.

What do you get at the end of a knowledge audit?

Three things worth having. The first is a written map of one or two real processes end to end, including the waiting and the rework that nobody usually counts. The second is a prioritised shortlist of candidate projects with a rough scope and an estimate of the hours each would free, and the third is an inventory of what your business knows that is currently undocumented, with a view on which parts are worth writing down first. All of it should be yours to keep, in a format you can read and hand to anyone, including a different supplier. If what you get back is a slide deck of recommendations with no underlying map, you have bought a sales document.

How long does a knowledge audit take?

One to two weeks of elapsed time for a focused audit, and the demand on your team is smaller than people expect: usually a few conversations of an hour each with the people who actually do the work, plus access to look at how things currently run. Adding a phased roadmap on top takes it to around two to three weeks. Anything being sold as a multi-month discovery programme before a single thing gets built is a warning sign rather than a sign of thoroughness, because the diagnosis stops being useful once it is older than the process it describes.

What does an AI readiness assessment cost in the UK?

For a focused audit of one or two processes, expect to start from around £3,000. We do not take engagements much below that, because the time cost of discovery and onboarding makes them uneconomic on both sides. If the audit runs straight into delivery, a first build phase typically lands between £6,000 and £25,000 depending on how many systems have to be connected and how much of the reasoning is already written down. Those are the honest ranges for mid-market work. If you are being quoted six figures to be told whether you are ready, you are buying a consultancy process rather than a diagnosis.

Do we need an audit before building anything?

Not always, and it is worth saying so plainly. If you already know exactly which task is eating the hours, and that task sits in one place with a stable process behind it, skip the audit and scope the build. The five readiness questions in our capability-plan post will get you most of the way for nothing. An audit earns its money when the opposite is true: several candidate problems and no agreement on which is worst, work that crosses departments so nobody sees the whole path, or a previous attempt that stalled and nobody can say why. Paying to find out where the problem is only makes sense when you genuinely do not know.