If you believe the vendor blogs, AI consulting is where the money is. The UK contract market disagrees, and it has 3,018 quoted day rates to make the point.
The median AI contract day rate is £559. A year ago it was £550. That is a rise of 1.64%, which after inflation is a real-terms cut. Meanwhile data science, a discipline that has been unglamorous since roughly 2019, rose 12.94% to £600.
The most-hyped skill in the market is the flattest-priced skill in the market.
AI contract rates, August 2026
Median advertised daily rates on the UK contract and interim market for AI, machine learning and data science roles.
- Source
- IT Jobs Watch, which derives medians from daily rates quoted in advertised UK contract vacancies. Per-skill sample sizes are shown in the table below.
- Window
- The six months ending 21 August 2026, compared against the six months ending 21 August 2025.
- What this measures
- What an individual contractor is paid for AI work. It is not what a consultancy bills a client for an AI engagement, which is covered separately below.
- US comparison
- US figures are federal ceiling rates from GSA CALC+, pulled 28 September 2026: the most a contractor may charge a US federal agency for that labour category. They are not contractor pay, and they are not what a firm quotes a commercial client.
- Review cycle
- Reviewed quarterly against the same source so year-on-year comparisons stay like for like.
ConsultingDemand, “AI Consulting Rates in 2026”, data as of 21 August 2026. https://consultingdemand.com/blog/ai-consulting-rates/
The three AI rates, and why they differ
“AI consulting” is not one job. The contract market prices at least three of them differently.
| Skill | Median day rate | Year ago | Change | Sample |
|---|---|---|---|---|
| Data Scientist | £600 | £531 | +12.94% | 217 |
| Machine Learning (skill) | £575 | £556 | +3.42% | 1,018 |
| Artificial Intelligence (skill) | £559 | £550 | +1.64% | 3,018 |
Artificial Intelligence and Machine Learning are skill markers, not job titles: IT Jobs Watch counts vacancies citing the skill across every contract role, so they overlap the job titles above and cannot be added to them.
Look at the sample column, because it explains the rate column.
Artificial Intelligence appears in 3,018 quoted contracts. That breadth is exactly why the rate is flat. When a skill appears in three thousand adverts, it has stopped being a specialism and become a keyword. Plenty of those roles are ordinary engineering or analysis work with “AI” in the description because the hiring manager knows it attracts applicants.
Data Scientist appears in 217. It is a narrower, better-defined role with a harder skills test, and it is the one that gained 13%.
The pattern is consistent: the more precisely a market can define what it is buying, the more it will pay for it.
The gap between contractor pay and consultancy billing
This is where the vendor blogs get their big numbers, and they are not wrong so much as comparing different things.
Two different purchases, two different numbers
The public US benchmark is federal ceiling rates: the most a contractor may charge a US federal agency for that labour category. For AI and data science roles the middle half runs $138 to $208 an hour, median $171, across 1,602 rates in GSA CALC+ in September 2026 (consulting fees by industry). That is the highest median of any industry on the schedules. At a notional eight-hour day the median comes to $1,368 (our arithmetic: $171 × 8), for a firm billing the work rather than a contractor paid for it.
Both numbers are true. They describe different purchases. If a firm quotes you its rate and you benchmark it against the contractor figure, you will conclude you are being fleeced, and you will be wrong. The reverse error is more expensive: hiring a contractor at £559 and expecting consultancy-grade delivery assurance that nobody is being paid to provide.
Ask who is accountable if the model underperforms in production. A contractor is accountable for their time. A consultancy is accountable for the outcome, and the difference in price is largely the price of that accountability. If a firm charges consultancy rates but writes contractor-grade accountability into the contract, you are paying for something you are not getting.
What an AI engagement actually costs to buy
Day rates only help if you already know how many days you need. For scoping a budget, engagement shapes are more useful, and public prices for them are scarce.
On the UK public record, five AI strategy and readiness contracts from 2024 to 2026 ran from £27,000 for an AI roadmap to £104,000 for a six-month advisory assignment; our UK consulting day rates guide lists them with links. We found no published source for pilot or programme fees, so there are none on this page. Get each phase quoted separately: the build that follows the assessment is a different SoW, and it should not be assumed in the first one.
The pattern worth noticing is that the expensive part is never the model. It is integration, data plumbing, change management, and the assurance work that lets a regulated business put the thing in front of customers. Firms that quote cheaply for a proof of concept and expensively for production are not upselling. They are pricing where the work actually is.
Why flat rates do not mean flat demand
It would be easy to read 1.64% as the AI bubble deflating. The contract volume argues otherwise: 3,018 quoted rates is an enormous number of advertised contracts, far more than any other specialism in our benchmark.
Demand is not the constraint. Supply caught up. Two years of retraining, certification and career-switching produced a large pool of people who can credibly answer AI questions in an interview, and a large pool is a competitive one. Rates flatten when supply meets demand, even while both are growing.
The specialisms still rising are the ones where supply has not caught up, because the barrier is not a course you can take. Cyber security consulting rates rose 21.75% for exactly that reason.
Data science's +12.94% comes from 217 quoted rates and AI's +1.64% from 3,018. The AI figure is the most statistically robust number in our entire rate benchmark, and it is the one showing least movement. When a widely-sampled figure is flat and a narrowly-sampled one spikes, trust the flat one.
If you are buying
Three things worth doing before you sign anything.
Ask what they have put into production. Not what they have built. What is running, for whom, and for how long. The gap between a working demo and a system a regulated business depends on is where most AI budgets die. Our guide to hiring an AI consultant has the full question set.
Price the integration separately. If the proposal bundles model work and integration into one number, you cannot tell which part is being underestimated. It is almost always the integration.
Benchmark against the right market. Contractor or consultancy, pick one and compare like with like. Consulting fees by industry sets out US federal ceiling rates and public project fees across twelve industries; the table above covers UK contractors.
- UK AI contract day rates sit at £559, up only 1.64% year on year, which is a real-terms cut after inflation
- Data science rose 12.94% to £600 in the same window, so the AI-adjacent premium is real but it is not attached to the word 'AI'
- The AI figure draws on 3,018 quoted rates, making it the most robust number in the benchmark and also the flattest
- Flat rates reflect supply catching up, not demand falling — contract volume for AI skills is the highest of any specialism tracked
- UK contractor day rates (£559) and US federal ceiling rates for AI and data science roles ($138–$208 an hour) are different markets and should never be benchmarked against each other
- The expensive part of an AI engagement is integration and assurance, not the model, so get the production build quoted separately from the proof of concept
Last updated: 29 September 2026