What an AI consultant costs in 2026, in the UK and the US

Search for what an AI consultant costs and most of what comes back is a range with no source behind it. The ranges are wide enough to be useless, and nobody says what the money buys. This post does it the other way round: what an AI consultant or AI consultancy costs in the UK and the US in 2026, split by pricing model, with every figure tied to a named source and a date, and an engineer's view of where the money actually goes.

The agency figures come from our own survey of 20 published price lists, read off each firm's website in August 2026 with no call booked. The full table is in AI automation agency pricing 2026, and the maintained version is the agency price index. Prices change, so check the firm's own page before you quote any of these back to someone.

Day rates: what the market pays one person

The cleanest independent number in the UK is the contract rate. IT Jobs Watch put the median AI consultant contract day rate at £564 in the six months to 18 October 2026, up from £544 a year earlier. The 10th percentile was £425 and the 90th £713, across 107 contract ads. That is what UK companies advertise to pay an individual contractor, usually through an agency or umbrella company, for work they direct themselves.

Consultancies sit above that, which is what you would expect, because the day rate has to carry sales, unbilled time, insurance and the person who covers when the first one is ill. Published UK day rates in August 2026 ran from £550 at Axivon and £800 at MQLFlow to £950 to £1,500 at The AI Consultancy. Winder.AI published £150 to £400 an hour, and Devaland a day rate from $750.

We did not find a US equivalent of IT Jobs Watch that we would trust, so the US picture is the published hourly rates: $150 at CloudNSite, $250 at XRAY, with prepaid blocks of $1,000 for 4 hours up to $5,000 for 20, and a blended $150 to $200 at Aloa. At eight hours a day, $250 an hour is a $2,000 day.

What a day rate pays for is time, and only time. It does not pay for an outcome. If the job takes twice as many days as anyone expected, and with real business data it often does, you pay twice. Day rates are fine for advisory work where you decide how many days you want, such as a review of a vendor contract or a second opinion on an architecture. They are a poor way to buy a build.

Paid discovery and audits: what you pay to be told what to build

Most consultancies now charge for the first look. In August 2026 the published fees landed on a few anchors: £495 at The AI Consultancy and Axivon, $999 at CloudNSite, £1,500 at The Automation Agency, $2,000 for consulting discovery and $2,500 for a single-workflow audit at Palavir, $2,500 at Justin McKelvey, and £3,500 for The AI Consultancy's two-week Readiness Sprint. At the enterprise end, Blue Orange Digital charged $25,000 to $50,000 to assess, and Winder.AI published strategy sprints at £10,000 to £50,000.

A paid audit should buy a document you can act on without the firm that wrote it: a map of how the work runs today, a ranked list of what is worth automating with the arithmetic shown, and a fixed price for the first build. What a paid AI opportunity audit should contain goes through it item by item. The detail to read twice is the credit. Some firms credit the fee in full against a build within a deadline, some credit half, some only if you go ahead, and some not at all. The difference matters most when the honest answer is that you should not build anything.

Free first consultations: what you get for nothing

Plenty of firms start with a free call. In August 2026 Zaps Studio offered a free 30-minute audit, LeftClick a free funnel audit, Justin McKelvey a free 20-minute repo audit, and Podlevskikh a free diagnosis within 24 hours.

A free call is worth exactly what you hold at the end of it. If it ends in a written proposal with a fixed price, or a plain statement that automation would not pay back for you, it did a real job. If it ends in an invitation to a paid discovery workshop, it was a sales meeting. Ask before you book.

Fixed-price projects: what you pay for an outcome

This is where most of the money in a 10 to 250 person business should go, because it is the only model where the price is attached to something that runs. Published mid-market fixed builds in August 2026 started at £350 for one simple automation at The Automation Agency and ran up through builds from $8,000 at CloudNSite, $15,000 for the top fixed SKU at Startrise AI Labs, and $25,000 for Palavir's ten-day build. Above that you are buying a different kind of firm: Aloa quoted $50,000 to $150,000 for production-ready work, and Winder.AI £120,000 to £400,000 for mid-market production.

A fixed price pays for a written scope, the build, and, if it is any good, the hardening: error handling, retry limits, logging, an alert that reaches a person, and testing against your real data rather than a tidy sample. In our own work that hardening is roughly as much effort as the first working version, and it is the part a cheap quote leaves out. The case for fixed-price AI work covers why we prefer it.

Retainers: what you pay for someone to stay

Retainers on the published lists ran from £200 a month at MQLFlow and $300 at Podlevskikh to £3,000, £5,000 or £8,500 for The AI Consultancy's fractional Chief AI Officer, $7,500 to $10,000 for Palavir's consulting retainer, and $15,000 to $30,000 for Blue Orange Digital's advisor tier.

That spread hides two different products. An advisory retainer buys a senior person's attention for a few hours a week: meetings, reviews, a roadmap kept up to date. A care retainer buys somebody watching systems that already exist, fixing them inside an agreed window, and improving one thing a month. Both are legitimate. They are not interchangeable, and a business with no running systems has nothing for a care retainer to look after. What a care retainer should include lists what the second kind should contain, and what a fractional AI engineer costs covers the version where the person writes code.

Where the money goes

Whatever the model, the bill for AI consulting is mostly people's time, split across five things:

1. Finding out how the work really runs. Interviews, reading the systems, sitting next to the person who does the job by hand. Skipping this is how a build ends up automating the process as described rather than the process as it happens.
2. Writing the scope. What done means, which systems are touched, which edge cases are handled, and the acceptance test.
3. The build. Often the smallest of the five.
4. Hardening and testing. The part that decides whether it still works in month three.
5. Running it. Model usage, platform fees and someone's time after handover. Winder.AI publishes year-one operations at 15 to 25% of the build cost, a useful rule of thumb even when your supplier will not say it. The running costs nobody mentions breaks that bill down.

The AI model itself is rarely a large line. If a quote is high, it is high because of the people, and you should be able to see which of the five they are spending on.

How to compare two quotes

Put each quote into the same shape before you look at the totals:

• What do I hold at the end? A document, a prototype, or a system running in production on my data.
• Is the price fixed or a rate? If it is a rate, is there a cap, and what happens at the cap?
• Who does the work? The person on the sales call, or someone you have not met.
• What is credited? Any paid discovery, against what, and by when.
• Whose accounts? Everything should be built in accounts in your company's name.
• What does month two cost? Warranty, support window, and the monthly running bill.

Two quotes that answer those six the same way can be compared on price. Two that answer them differently are not two prices for the same thing.

Red flags

• No price, range or floor anywhere until you have booked a call. Expect the number to be sized partly to your budget.
• A guaranteed ROI percentage. Nobody can know that before they have seen your process.
• Paid discovery with no defined deliverable and no credit.
• A day rate with no estimate and no cap on a build.
• A twelve-month retainer before anything has shipped.
• A strategy deliverable that names no specific process, system or price.
• Accounts, code or prompts held in the consultancy's name.

Ten questions before hiring an AI agency turns these into things to ask on the first call.

Our own prices, for reference

So there is one fixed point to compare against: we start with a free 30-minute consultation that ends in a written proposal with a fixed price within two business days. A written diagnostic for a broken system is from $500 (£400) and credited in full against the repair. Build sprints are $3,000 to $25,000 (£2,300 to £19,500), care plans $1,000, $1,800 or $3,000 a month (£800, £1,400 or £2,300), and a fractional engineer $6,000 to $12,000 a month (£4,700 to £9,400). The full list, with what each includes, is on the pricing page.