Hire in-house when AI is becoming a permanent, differentiating capability; hire an AI consultant when you need speed, breadth of pattern recognition, or a defined outcome this quarter. Most UK mid-market companies get the best economics from a third option: a short consultancy engagement that trains internal champions, then a sharper in-house hire six months later. If you land on the consultancy route, our comparison of the best AI consulting firms in London and the UK covers who to shortlist; this page covers the decision before that one.
Key Takeaways
- An in-house AI lead in London typically advertises at £90,000–£150,000 base — £110,000–£190,000 in real annual cost once employer's NI, pension and tooling land — plus three to six months to hire and ramp.
- A consultancy engagement runs £3,500–£45,000+ for a defined outcome starting within weeks. The comparison is not salary vs fees; it's cost per outcome per quarter, including the outcomes that never arrive while a role sits vacant.
- One hire cannot transform an organisation: a lone AI manager without mandate, budget or executive air-cover becomes a help desk. The failure is structural, not personal.
- Consultancies are the wrong answer for permanently embedded capability: if AI decisions will be made weekly forever, renting the decision-maker indefinitely is the expensive version.
- The hybrid sequence — audit, champions trained, then hire with a written role scope — beats both pure options on evidence: you hire later, cheaper mistakes, and the new hire inherits usage data instead of a blank page.
The real cost of each option
| In-house AI lead / Head of AI | AI consultancy | |
|---|---|---|
| Headline cost | £90,000–£150,000 base (London advertised range); more at enterprise Head of AI level | £3,500 (audit) to £45,000+ (quarter-long programme) |
| Real cost, year one | £110,000–£190,000+ with employer's NI, pension, tooling, recruitment fees | The fee, plus your teams' engagement time |
| Time to first outcome | 3–6 months to hire, then ramp | 2–6 weeks |
| Breadth of pattern-matching | One person's career | Cross-company: what worked and failed across dozens of firms |
| Permanence | Compounds — if the org supports the role | Ends by design; value persists only if trained in |
| Failure mode | Isolated hire becomes internal help desk | Deck-ware if no enablement was bought |
Salary figures reflect typical advertised London ranges in 2026; regional roles run 10–25% lower. Consultancy bands are unpacked line by line in our UK AI consulting cost guide.
When the in-house hire is right
- AI is becoming core product or infrastructure. If models, agents or data products are your differentiation, you need owners on payroll, and probably a team.
- Decision volume is permanent. Weekly calls on tooling, vendors, governance and use cases, indefinitely — renting that forever costs more than employing it.
- You've already proven demand. Usage is real, the roadmap is full, and the constraint is dedicated ownership rather than knowledge.
When the consultant is right
- The problem is a project, not a function. A readiness audit, a governance framework, a training programme — defined start, defined end. What each engagement type should cost is documented here.
- You need the cross-company sample. A good consultancy has watched the same rollout succeed and fail across dozens of companies; a first AI hire has seen their previous employer.
- Speed dominates. Weeks matter, and a six-month recruitment cycle is itself the risk.
- You don't yet know what the job is. Writing the Head of AI job description is an output of diagnosis — see advisory vs implementation vs training — not a prerequisite for it.
"The first AI hire fails for structural reasons: no mandate, no budget line, no executive sponsor, and a job description written before anyone knew what the job was. Six months later a capable person is answering prompt questions on Slack. We're often hired to fix exactly this — which is absurd, because a £3,500 diagnosis before the hire would have prevented it."
— Meera Sanghvi, Co-Founder, Spicy Advisory
The hybrid most mid-market companies should run
- Weeks 1–3: diagnose small. A fixed-scope readiness audit maps value, skills and governance gaps. Cost: £3,500–£15,000.
- Quarter one: train champions, not a hire. An adoption training programme builds internal champions inside real workflows, with usage measured. The organisation learns what it actually needs.
- Month 4–6: hire against evidence. Now the role scope writes itself from usage data: which teams, which stack, which governance load. Recruit with the consultancy's scorecard, inherit the champions network on day one.
This sequence costs less than one mis-hire — recruiter fees plus six lost months — and it converts the consultancy's exit from a cliff into a handover. It's also, candidly, how we prefer to work: our engagements are scoped to make ourselves unnecessary, which is the position argued across our firm comparison.
"The question isn't consultant or hire — it's what evidence each pound buys. A hire before diagnosis buys you a guess with a twelve-month notice period. A quarter of measured adoption buys you the job description, the shortlist criteria and usually the internal candidate."
— Toni Dos Santos, Co-Founder, Spicy Advisory
The internal-candidate surprise: in a meaningful share of champion programmes, the right first AI lead turns out to already work for you — a operations or product manager who becomes the obvious owner once trained. Cheaper than recruitment, faster than onboarding, and already trusted by the teams.
Frequently asked questions
Do I need an AI consultant or an in-house hire first?
For most 50–1,000 person companies: consultant first, but scoped to end. Use a short audit and one quarter of champion training to generate the evidence, then hire in-house against that evidence. Hire first only if AI is already core to your product.
How much does a Head of AI cost in the UK?
London advertised bases typically run £90,000–£150,000 for AI lead roles and higher for enterprise Head of AI positions; add roughly 20–30% for employer costs, plus recruitment fees and a multi-month ramp. Budget the role, not just the salary.
Can a consultancy replace a Head of AI long-term?
No, and be wary of one that says yes. Fractional advisory works as a bridge, but permanent weekly decision-making belongs on payroll. The consultancy's job is to make the eventual hire smaller, later and better-informed.
What should the first in-house AI hire look like for a mid-market company?
Usually a translator, not a researcher: someone who turns workflows into use cases, owns governance day-to-day and runs the champions network. Hire builders once usage data proves which systems deserve them.
Sources
- UK advertised salary ranges for AI lead and Head of AI roles, London job boards, 2026 — ranges deliberately quoted wide; verify against live listings for your sector
- ONS and British Chambers of Commerce adoption data — via our UK SME AI statistics roundup
Buy the evidence, then decide
Our audit (£3,500) and 90-day programmes are scoped to answer exactly this question with your data — including, sometimes, "hire now, here's the job description". Founder-led, 50+ companies, 1,500+ professionals trained, 4.98/5. We'll tell you plainly which side of the table you need.
Run the Free AI Adoption Scorecard Book a Free 20-Minute Call