When should you hire a permanent AI lead instead of a fractional one?
6 minute read. Updated 2026-08-08.
Hire a permanent AI lead once the role needs 3 days a week or more, sustained for longer than 2 quarters. Below that, a fractional hire is cheaper and more flexible. At 3 days a week, using the advertised median contract rate of £551 a day, the annual cost already exceeds the advertised median permanent salary of £71,000, before adding a single employment cost.
The rule
Hire a permanent AI lead once the role needs 3 days a week or more of dedicated attention, sustained for longer than 2 quarters. Below that threshold, or for a defined project with an end date inside 2 quarters, a fractional hire wins on both cost and flexibility.
Everything below is the arithmetic and the exceptions, because a rule without arithmetic is just an opinion, and opinions about headcount rarely survive a finance review.
The arithmetic at 3 days a week
Take the advertised median contract day rate for artificial intelligence roles broadly, £551, over the six months to August 2026. At 3 days a week for 50 weeks, that is 150 days a year: 150 x £551 = £82,650.
Compare that with the advertised median permanent salary for the same broad category, £71,000. Even before adding employer National Insurance or pension contributions to the permanent side, and before accounting for the gaps a fractional lead needs between engagements, the contract route already costs more at 3 days a week.
The permanent salary figure also excludes pension, bonus and equity, so the true cost of the permanent hire is higher than £71,000. It is still, at this volume of days, the cheaper of the two.
Where the arithmetic reverses
Drop to 1 day a week and the picture flips completely. 50 days a year at £551 is £27,550, a fraction of any permanent salary, and you avoid the fixed costs of employment: recruitment, onboarding, notice periods, and the risk of hiring the wrong permanent person into a role that did not need to exist full time.
Most companies buying AI leadership for the first time genuinely need something closer to 1 or 2 days a week: someone to set direction, review what a build team is doing, and stop obviously bad decisions before they ship. That is a fractional job, and paying a permanent salary for it wastes money for months before anyone notices.
Why duration matters as much as volume
Volume alone does not settle the question. A 3-month project that genuinely needs 4 days a week, such as an initial model risk review or a first production deployment, is still better run fractionally, because the need has a visible end date. Hiring permanent for a role that stops mattering in 3 months leaves you carrying a salary with nothing left for the person to lead.
The 2-quarter mark matters because that is roughly where the cost of repeated fractional engagements, plus the cost of re-explaining context to a new person if your original contractor moves on, starts to exceed the cost of continuity. Institutional knowledge in AI work compounds: the person who knows why a model was built a certain way 6 months ago is worth more than the person who is simply available this month.
IR35 and the exclusivity problem
If your engagement is determined to fall inside IR35, deductions through an umbrella company substantially reduce what the contractor actually takes home, and contractors who know this tend to price sustained, exclusive roles higher to compensate. Ask a fractional lead to be in your building 4 or 5 days a week for a year and you are functionally asking for an employee's exclusivity while still paying a rate built around a contractor's flexibility.
That mismatch is not a legal problem you can fix with better contract wording. It is a sign the role has become a permanent one in every respect except the payroll line, and it should move onto the payroll line.
What changes at the senior end
At the top of the market the gap widens further. The advertised 90th percentile contract day rate for artificial intelligence roles is £788, against a 90th percentile permanent salary of £111,250. Advertised figures are not agreed figures, and that gap between what is asked and what is actually paid tends to widen the further up the range you go.
For a genuinely senior lead, sustained past the 2-quarter mark, the arithmetic and the continuity argument both point the same way. Permanent is very likely to be both cheaper and better once you are past 3 days a week.
Test the rule against your own numbers
Take your honestly expected weekly need over the coming 2 quarters. Multiply the day count by the median contract rate for the closest role category in the table: machine learning engineer at £575, data scientist at £600, data engineer at £500, MLOps at £575. Compare the result with the median permanent salary for the same category.
If the contract total already exceeds the permanent salary before you have added a single employment cost, you are past the threshold. Stop treating the fractional hire as the safe default and start writing the permanent job description.
What to do about it
- Work out your honestly expected days per week before choosing between permanent and fractional.
- Treat 3 days a week, sustained beyond 2 quarters, as the point where permanent usually wins on cost.
- Do not hire permanent for a role with a visible end date inside 2 quarters.
- Remember permanent salary figures exclude pension, bonus and equity, so add those before comparing.
- Watch for a fractional lead being asked for near-full-time exclusivity; that is a sign the role should move onto payroll.
- Recalculate using the day rate for the closest specific role, not the broad AI figure, once you know what you actually need.
Questions people also ask
What if I only need someone for 3 months but at 5 days a week?
Short, intense, time-boxed work such as an initial model risk review or a first deployment still suits a fractional hire, even at 5 days a week, because the engagement has a visible end. The 2-quarter threshold is about sustained need, not raw days per week; a defined project with an end date is a project, not a permanent role.
Does the arithmetic change for a specialist role like MLOps rather than a general AI lead?
Yes. Use the median for the specific role rather than the broad AI figure. MLOps carries an advertised median contract rate of £575 a day and a median permanent salary of £87,500, both higher than the broad AI figures, so the crossover point shifts slightly but the same 3-day, 2-quarter logic applies.
Can I just hire fractional and convert to permanent later if the need grows?
Yes, and this is often the sensible order. Start fractional while you learn what the role actually needs to cover, then convert once the pattern is clear and sustained. Be honest with the contractor about this possibility early, since some will price differently if they know a permanent offer might follow.
Isn't a permanent hire always cheaper in the long run anyway?
No. Below roughly 2 days a week, a permanent salary is close to guaranteed waste, because you are paying for capacity nobody will use. The crossover in this piece exists precisely because the answer depends on volume and duration, not on a general belief that permanent is always the frugal choice.
Where the figures come from
Every rate and salary quoted in this article is a median or percentile of figures advertised in UK job postings over the six months to 8 August 2026. They are not rates paid, and the gap widens at the top of a range.
- IT Jobs Watch, UK contract rates, 6 months to 8 August 2026, read 2026-08-08.
- IT Jobs Watch, UK permanent salaries, 6 months to 8 August 2026, read 2026-08-08.