You are nine months into AI adoption. You have worked with two consultants, one good, one less so. Your CFO has run the numbers and asks why you are not just hiring an AI lead. It looks cheaper on paper. You ask the question to a peer, then to your accountant, then to LinkedIn. The answers are inconsistent.
This is one of the genuinely hard decisions in AI capability building at SME scale, and the answer depends on the shape of the work, not the preference for in-house or consulting. The maths comes out one way for some businesses and the other way for others, and the deciding factor is the volume and continuity of the AI work itself.
Why is this question structural, not preferential?
The in-house versus consulting question sounds like a culture decision but resolves like a capacity decision. If you have continuous AI work for one full-time person, in-house is cheaper and better. If you have episodic AI work that comes in bursts of one to two months at a time, consulting is cheaper and better. The mistake most SMEs make is treating it as a preference question, then discovering eight months in that they hired for a load they did not have.
The deciding question is how much continuous AI work the business will actually have once the initial rollout is done.
What an in-house AI hire actually costs in year one
Salary expectations for AI roles at UK SME scale run £45,000 to £85,000 for mid-level talent and £80,000 to £130,000 for senior leads. Add 25 to 35% for employer on-costs (national insurance, pension, equipment, training, recruitment), and another 30 to 50% loading on year one for ramp-up productivity loss while the new hire builds context. The total first-year cost typically lands £56,000 to £176,000 depending on seniority.
That number is materially higher than the headline salary, and it is the right number to compare against consulting. A £75,000 mid-level AI lead, with £20,000 of on-costs and a 40% ramp-up loading on year one, costs £133,000 in the first 12 months before they have shipped a deliverable. The same £133,000 buys roughly 165 to 220 days of senior consulting at UK rates, which is enough to run two mid-sized AI engagements and a strategic retainer.
The break-even is not abstract. It sits at the point where the in-house person’s loaded cost falls below what the same delivery would have cost in consulting fees, and that point depends entirely on how busy the in-house person actually is.
Where the break-even sits
The break-even with consulting typically falls at 3 to 4 concurrent mid-sized AI projects running continuously across the year. Below that, the in-house lead has more capacity than work. Above it, consulting becomes unwieldy to manage and in-house starts to win on cost as well as continuity.
A mid-sized AI project for an SME absorbs 30 to 60 days of senior delivery time, plus ongoing support. Three concurrent projects through a year is roughly 200 to 300 days of work, which is what one full-time AI lead can deliver. Below 200 days a year, you are paying for unused capacity. Above 300 days, you are paying premiums to consultants who could have been one full-time hire.
Most SMEs in the first 12 to 18 months of AI adoption sit well below the break-even. One to two AI initiatives running at a time, each taking 30 to 60 days, totalling 80 to 120 days a year of senior delivery time. That is part-time consulting territory or a fractional CAIO retainer, and an in-house hire would be sitting idle 40 to 60% of the year.
What in-house gives you that consulting cannot
In-house wins on three things that matter when AI is becoming infrastructure rather than a project. Continuous availability. Organisational knowledge. Talent retention. None of these can be bought from a consultant at any rate.
A consultant turning up two days a week cannot match the trust an embedded employee builds with the team over six months. The embedded person knows which client the operations lead is worried about, which integration breaks at month-end, and where the data model is slightly wrong. That contextual knowledge compounds, and at the point where AI is genuinely woven into operations rather than added as a layer, the embedded person is the one who keeps it healthy.
Talent retention is the third advantage. A senior AI hire who is well-deployed becomes harder to replace and more valuable as the company grows. A consultant relationship can move to another client without warning. For a business where AI capability is becoming a strategic asset, the retention advantage matters.
What consulting gives you that in-house cannot
Consulting wins on three things that matter when AI work is bursty rather than continuous. Breadth of experience across many businesses. Exposure to current practice as the field moves. Faster cycle time when work appears, because the consultant has done the same shape of project before.
An in-house AI lead at £130,000 has, by definition, only one client’s experience. A consultant at the same loaded cost has seen 20 or 30 similar engagements, which means they have already encountered the patterns the in-house person will hit for the first time. For an SME without a massive AI roadmap, the consultant’s pattern recognition is worth more than the in-house person’s continuous presence.
Cycle time is the second advantage. A consultant can start a new piece of work the week you sign. An in-house hire is at 40% effectiveness for the first six months. If the work is episodic, the consultant ships before the hire is fully ramped.
The hybrid model that usually wins
Most SMEs at the awkward middle stage end up with a hybrid arrangement that wins on cost and capability. An episodic consultant for strategy, architecture, and new pilot work. A part-time in-house data steward or operations lead who maintains the deployed AI capability and keeps the tooling stack disciplined. The consultant returns for the next strategic question.
The economics work because you are paying for the strategic thinking only when you need it and paying for the operational availability at a level that matches the actual day-to-day work. A 0.5 FTE internal owner at £40,000 to £55,000 fully loaded plus a fractional consultant at £30,000 to £60,000 a year covers most SME AI rollouts in their first three years, and it costs less than either pure option.
The hybrid is unfashionable because neither side of the debate sells it. Consultants prefer to sell consulting. Recruiters prefer to sell hires. The hybrid is the answer most SMEs land on after they have tried one of the pure options and found it expensive.
Which path is right for you?
Two questions decide it. How much continuous AI work will your business have once the initial rollout is done? And how much of that work needs to be done by someone who knows your business deeply rather than someone who has seen 20 similar businesses?
If the answer to the first question is “less than two FTE-equivalent of work a year” and the answer to the second is “the strategic part, not the operational part”, you are in consulting or hybrid territory. If continuous work is closer to a full FTE and the operational depth matters more than the strategic breadth, you are in in-house territory. The maths is structural, and the answer is rarely the one the CFO’s spreadsheet shows on first pass.
If you would like to talk through which path fits your business, book a conversation.
