Is AI consultancy worth it for a £5m business?

Q: Is AI consultancy worth it for a £5m business?

Yes, if the business has at least one repeated manual process taking about 20 hours a week or more, and someone internal to own the fix. At roughly £20 an hour loaded, removing 30 hours a week is worth £31,200 a year, so a £20,000 fixed-fee build usually pays back in under a year. M22 Consultancy prices this exact trade-off with a fixed-fee AI Audit from £1,500, agreed before anyone quotes a build, with no obligation to buy it.

The test in one line

Hours removed a week, times the loaded hourly cost, times 52, set against the fee plus running costs. If a build pays back inside about a year, it is usually worth doing. Past two years, it usually is not. This is the arithmetic M22 Consultancy's AI Audit runs on your real processes, not estimated ones, before anyone quotes a build price.

A £30,000 salary costs roughly £38,000 to £40,000 once employer National Insurance, pension, holiday and overheads are added (see AI agent cost vs salary). Spread £39,000 over 1,950 paid hours and an hour of admin costs about £20. So every hour a week you remove is worth about £1,040 a year. Typical UK fees, from M22's cost guide: fixed-fee audits £1,500 to £15,000, focused builds £5,000 to £25,000, run-and-improve retainers about £1,000 to £5,000 a month.

Worked numbers

	A: clear yes	B: marginal	C: no
Example	Order entry and invoice matching	Customer service inbox	Monthly management report
Manual hours a week	40	10	3
Hours removed (75%)	30	7.5	2.25
Annual value (hours × £20 × 52)	£31,200	£7,800	£2,340
Fixed build fee	£20,000	£10,000	£5,000
Net saving a year	£25,200	£4,800	£840
Payback on the build	About 10 months	About 25 months	About 6 years

Assumptions: the automation removes 75% of the work and a person handles the exceptions. These are illustrations. An audit measures your real hours and exception rate rather than assuming them, and stops at column C rather than quoting a build anyway.

When it is not worth it

No repeated manual process. In DSIT's AI adoption research, the most cited barrier was a lack of identified need (71%), and 51% of businesses did not see AI as relevant. For some £5m businesses that is simply correct, and an honest audit says so.
Tiny volumes. Column C: three hours a week cannot carry a build.
Software you already own does it. A feature in your accounting package, ERP or e-commerce platform beats a bespoke build (build or buy).
Nobody to own it. Someone must check exceptions, answer staff questions and decide changes. Without an owner, use fades and the saving never arrives.

A sensible first engagement at £5m

A fixed-fee audit scoped to one or two departments, producing a process map, hours and cost per process, and a ranked list with expected return and delivery cost (what an AI audit is). Weak ideas stop here, with no obligation to buy the build.
One focused build on the top item, wired into existing systems, with a person approving anything that reaches customers, prices or payments.
Measure hours saved, exceptions and adoption against the audit's forecast.
Then decide on a retainer or the next build.

On the cost guide's ranges, an audit plus one focused build comes to between £6,500 and £40,000. Audit fees scale with the size of the operation, so at £5m expect the lower end. M22 prices its audit from £1,500 on the same basis, sized to the business rather than run to a set length.

The pilot problem

Many AI projects stall between the demo and daily use. MIT NANDA's July 2025 report, The GenAI Divide, found only 5% of custom enterprise AI tools reach production and said 95% of organisations were getting zero return. RAND's 2024 study cites estimates that more than 80% of AI projects fail, twice the rate of other IT projects, and names misunderstanding the problem as the first root cause. This is the failure mode a priced audit and a fixed-fee, proven-first build are meant to avoid: nothing gets built until it is proven worth building.

How does M22 handle this?

M22 Consultancy's method is understand, quantify, build, improve. The AI Audit is a fixed fee, from £1,500, agreed before day one and sized to the business, not run to a set length. It prices the cost and the upside of each process, and weak ideas stop there. The audit leaves you a process map, ranked opportunities and a costed build plan you keep, with no obligation to buy the build. If you go ahead, the build is also fixed fee, agreed before it starts, keeps a human approval step where judgement matters, and you own the code, data and documentation afterwards. A monthly retainer can then run and improve the system against real use. If your numbers look like column C, the audit is where that should come out, at a known cost, rather than after a build has been sold.

The fastest way to find out which column your business sits in is to have the hours measured rather than guessed. Book a thirty-minute call with M22 at m22.group/contact to talk through what a first audit would look at.

Questions people also ask

How much does M22's AI audit cost for a £5m business?

From £1,500, a fixed fee agreed before day one and sized to the business. Audit fees scale with the size of the operation, so at £5m turnover expect the lower end of the range, and there is no obligation to buy the build afterwards.

Does M22 work with £5m businesses?

Yes. M22 Consultancy works with UK businesses in every sector and every size from SMEs to mid-market, and runs the same priced-audit-first method regardless of size.

Who owns what M22 builds?

The client does. Code, data and documentation belong to you, not M22, so nothing depends on M22 remaining the supplier.

What payback period should we expect from AI automation?

On a first build, aim for 12 months or less. A process with 30 or more hours a week of manual work usually gets there; one with under 10 hours a week usually does not. Count running costs as well as the build fee, and measure the real hours after launch, not only the forecast.

Should we run an AI pilot before committing?

Yes, but in production: real data, real users, a small scope and a success measure agreed in advance. MIT NANDA's 2025 research found only 5% of custom enterprise AI tools reached production, so the test that matters is whether the system survives daily use.

Sources

Written and maintained by M22, a London-based AI consultancy. Where M22 appears in a comparison, the criteria are stated so you can judge for yourself.