Every figure on this page was read from the supplier’s own public page on 21 September 2026 and is linked in Sources. Where a number is ours rather than the market’s, it says so.

At a glance

  • £500 to £750,000 and above is the full published UK range for 2026, and the spread is almost entirely a spread in scope.
  • Four phases carry the cost: readiness, discovery, build, and the run rate after go-live. Most proposals price the third one only.
  • Our own numbers are fixed and published: a free 30-minute fit call, a £2,000 fourteen-day audit, then a scoped build price.
  • The audit fee is not credited against a build.
  • One system we built has priced a real issued job in public: 63 measured lines, £98,214.72, every line open to inspection.

What does AI implementation cost in the UK?

There is no single number, and anybody who gives you one before reading your process is pricing their assumptions. What there is instead is a published market, and it is wider than most buyers expect.

At the small end, AI Workforce publishes £500 to £2,000 for a simple workflow and £3,000 to £10,000 for a mid-range build. Chris Garlick publishes three tiers on his own page: a focused fix from £500, a build at £2,000 to £15,000, and a retainer at £300 to £1,500 a month. Sharp Code puts its range in its own title tag at £15,000 to £250,000. SoftBlues describes the whole UK integration market at £15,000 to £750,000 and above, with most smaller firms budgeting £5,000 to £30,000 for a first automation. Helium42 publishes £25,000 to £150,000 for a typical small or mid-sized project, with consultancy day rates of £500 to £2,500.

Read those together and the market says something useful. A single, bounded piece of work that changes one process costs low thousands. A system that several teams depend on costs tens of thousands. A programme that touches the whole operation costs hundreds of thousands. Your job as a buyer is to work out which of those three you are actually commissioning, because most disappointing quotes are a supplier pricing one of them while the customer imagined another.

The four phases and what each one costs

An implementation is not one purchase. It is four, and they are usually bought from different budgets at different times, which is why the total surprises people.

Readiness is the work of finding out whether the problem is worth solving and whether your data can support solving it. It is the cheapest phase and the one most often skipped. Ours is fixed at £2,000 for fourteen days.

Discovery and pilot is where the chosen process gets mapped properly and a narrow version is tested against real data rather than a demo set. Square Root puts discovery and technical planning at 5 to 10 per cent of a total software budget, which is a reasonable planning assumption for AI work too.

Build is the phase everybody quotes. It is also the phase whose price moves least once the first two have been done honestly, because by then there is something specific to price.

Run rate is hosting, model usage, monitoring and somebody to call. Square Root puts ongoing maintenance and support at 15 to 25 per cent of build cost annually, and cloud infrastructure at £200 to £2,000 a month. Published UK retainers run from £200 to £20,000 a month. If a proposal does not have a line for this, it is not finished.

What the audit fixes before anyone quotes a build

Two businesses ask for the same thing in the same words and need different software. One operations director at a 40-person distributor wants order acknowledgements read and filed. Another wants the same sentence to mean reconciliation against a live stock position with an exception queue. The words are identical. The builds are not.

A readiness phase resolves that before money moves. Ours produces four documents the client owns: a constraint map, an AI opportunity map, economic cases, and a ranked next step. Three sixty-minute working sessions with the engineers who would scope the build sit inside the fourteen days. The point of those documents is that they survive us. You can hand them to another supplier and get a comparable quote, which is the test of whether a readiness phase was real.

What is actually in a build quote?

A quote that only has one line on it is not a quote, it is a number. What you want to see is the shape of the work, because that is what tells you whether the supplier understood the job.

  • Integration work, named system by named system, with a note on what happens when each one changes at the other end.
  • Data preparation, with an explicit statement of what state the supplier believes your data is in and what they will do if they are wrong.
  • The decision rules, listed. Every rule is a branch that has to be specified, built and tested, and ten rules is more than ten times one rule because they interact.
  • Correctness work: validation, an audit trail, and a way for a person to catch a bad output before it leaves the building.
  • Handover: who owns the code, who owns the data, and what you would need to run it without the supplier.

Which costs never make it into the proposal?

Five of them recur, and every one of them is foreseeable.

Data preparation. The single largest overrun on most builds, and almost never in the customer’s original estimate.

Change of working practice. The system is finished and the team still does it the old way. This is not a software cost, which is exactly why nobody funds it.

Model and API usage. It scales with volume. A pilot on 200 documents a month says very little about a live system on 20,000.

The regulatory overlay. If the output affects an individual, the ICO’s guidance on AI and data protection applies to it, and working out what that means for your process takes time whether or not anybody budgeted for it.

The failure you do not see. An automation that silently stops is worse than one that never existed, because the work stops being done and nobody notices for a fortnight. Monitoring is a line item.

Why do we quote a fixed price and not a day rate?

A day rate is a real number, which makes it feel like solid ground. It tells you what an hour of someone’s attention costs. It does not tell you how many hours the job takes, and that is the whole question.

SoftBlues puts UK independents at £500 to £2,000 a day and the large firms between £2,000 and £5,000 a day. Helium42 publishes £150 to £400 an hour with typical projects running £15,000 to £60,000 over three to four months. Both are honest numbers and neither answers the buyer’s question. Two suppliers on the same day rate can differ several times over on the same brief, because one has built the thing before. Experience shows up as fewer days, not as a lower rate.

A fixed price moves the estimating risk to the person who is better placed to carry it. That only works if the scope is real, which is why we will not quote one before the audit.

What changes the number

Four things move it more than anything else.

The state of your data. One clean export is a morning. Five systems that half-agree with each other is the project.

How many systems it has to talk to. Each integration is a dependency you do not control, and each needs error handling for the day the other end changes without telling you.

What it costs when the output is wrong. A tool that suggests a task order can be wrong occasionally. A tool that issues priced quotes to customers cannot.

Whether the sector is regulated. Financial services, healthcare and anything touching employment decisions carry an approval path that is longer than the build.

A build we put into daily use, and what it produces

We will not price your build from this page. What we will do is show you what a piece of software we built actually produces, priced to the penny, so you can judge the standard before you judge a quote.

QSQuoter priced a first-floor build-over extension in Heaton Moor, Stockport, issued by Cheadle Construction on 14 July 2026 under reference JAY-AA-20260713. Sixty-three measured lines. £98,214.72 including VAT. Preliminaries and site set-up at £17,374, bathrooms and en-suites at £15,962, and eight more sections, all published. Open it and read every line. That is one specific claim, and it is the only kind we make: it shows we can build software that prices a construction job to an auditable standard. It does not prove hours saved, and we do not say it does. What we cannot show you yet is on our evidence page, stated plainly.

How does this compare with hiring in or buying a tool?

Three routes, and the right answer depends on how long the work lasts.

Buy a tool. Cheapest to start, fastest to prove, and the correct answer far more often than a supplier of custom work will tell you. A seat-priced product at £20 to £80 per user per month is hard to beat if it fits.

Hire an engineer. A permanent AI engineer in the UK is a salary plus employer costs plus the months before they ship anything, and one person cannot cover engineering, process design and adoption. It becomes the right answer when the work is permanent rather than a project.

Commission a build. Right when the process is genuinely yours, the tool market does not fit it, and the cost of the current way of working is several times the cost of changing it. That last clause is arithmetic, and the audit exists to do it on your figures.

When does the payback land, and when does it not?

Payback lands when three things are true: the process runs often, the time it takes is measurable, and somebody owns the new way of working. Miss any one and the payback is theoretical.

It does not land when the process runs monthly, when the saving is spread as five minutes across forty people, or when the number you are comparing against was never measured in the first place. A saving from a baseline nobody recorded is not a saving, it is a story. Put a loaded hourly cost against the hours the process actually consumes today, and if the total is comfortably under £2,000 a year, keep your money.

What this does not apply to

This page is written for UK businesses turning over roughly £1m to £50m that already have a process worth automating. It does not apply to sole traders or very small teams, where a subscription tool is nearly always the right answer. It does not cover public-sector procurement, which runs to different rules. It does not cover buying an AI product off the shelf, which is a different purchase with a different risk profile. And it does not cover research work, where the output is a finding rather than a system.

Every competitor band quoted here is that supplier’s published market commentary, not a quote for your work, and none of them is audited. We link them so you can check what we say they say.

When we are the wrong choice

If you want a proof of concept to show a board, we are the wrong people. We build things that run.

If you want a supplier who will quote from a one-page brief, we are the wrong people, and you should be careful of the ones who will.

If the honest answer to your problem is a £50 a month subscription, we will tell you that in the fit call and you will not owe us anything.

If nobody internally owns the process, no amount of software fixes it. Custom work makes a defined process faster. It does not create the definition.

Sources

Every figure above, with the page it came from. All read 21 September 2026.

If you want our numbers rather than the market’s, they are all on what it costs, including the worst case.

The short answers

Each answer is the opening of the section it links to, so nothing here is written for a crawler that a reader cannot also see.

How much does AI implementation cost in the UK in 2026?

There is no single number, and anybody who gives you one before reading your process is pricing their assumptions. What there is instead is a published market, and it is wider than most buyers expect.

What are the four phases of an AI implementation, and what does each cost?

An implementation is not one purchase. It is four, and they are usually bought from different budgets at different times, which is why the total surprises people.

Which costs are missing from most AI implementation proposals?

Five of them recur, and every one of them is foreseeable.

Why quote a fixed price for AI work instead of a day rate?

A day rate is a real number, which makes it feel like solid ground. It tells you what an hour of someone's attention costs. It does not tell you how many hours the job takes, and that is the whole question.

When does an AI implementation pay for itself?

Payback lands when three things are true: the process runs often, the time it takes is measurable, and somebody owns the new way of working. Miss any one and the payback is theoretical.

When is an AI implementation the wrong thing to buy?

If you want a proof of concept to show a board, we are the wrong people. We build things that run.

What is the cheapest AI implementation worth commissioning?

One workflow that runs often enough to be measured. Published UK entry points sit at about £500 to £2,000 for a single no-code automation and £1,500 to £5,000 for a supported build. If the process runs twice a month, none of those figures pay back, and the honest answer is to leave it alone.

What does AGMM charge for AI implementation?

The 30-minute fit call is free. The business-wide AI Constraint Audit is fixed at £2,000 and runs for fourteen days. A build is quoted as a fixed price only after that audit, because a number produced before anyone has read the process is a guess.

Is the £2,000 audit fee credited against a build?

No, and that is deliberate. If the fee came off the build price, we would earn more by recommending a build than by recommending you do nothing. Keeping it separate means the recommendation costs us the same either way.

Why do two quotes for the same AI project differ by five times?

Usually because they are quoting different things. One is pricing a pilot on clean sample data, the other is pricing a system that runs on your live data with error handling, permissions and an audit trail. Ask both suppliers what happens the first time the output is wrong.

What are the ongoing costs after an AI system goes live?

Hosting, model or API usage, and somebody to call. Published UK retainers run from £200 to £20,000 a month depending on how much of the system is yours to run. The variable that surprises people is model usage, because it scales with volume rather than with headcount.

Do we need to replace our existing software to implement AI?

Almost never. Most useful implementations sit beside the systems you already pay for and move work between them. Replacing a working system is the most expensive way to start and the slowest to show anything.

How long does an AI implementation take in a mid-sized UK business?

Published UK programmes describe eight to twelve weeks for a first production system and three to four months where several teams are involved. Our audit is fourteen days and produces a ranked, priced list before a build starts.

Can we implement AI without a data project first?

Sometimes. If the information the system needs already sits in one place and is reasonably clean, you can skip it. If it lives across a shared inbox, three spreadsheets and one person's memory, the data work is the project and pretending otherwise just moves the cost later.