For most SMEs and mid-market companies, hire a consultancy first and build in-house second. A capable in-house AI engineer costs £70,000 to £120,000 a year plus a six-month hiring and ramp period, and needs a pipeline of problems to justify the seat. A consultancy delivers a first working system in weeks on a fixed fee, and the right one leaves your team owning it, which is the moment an in-house hire starts to make sense.
The honest comparison
| AI consultancy | In-house team | |
|---|---|---|
| Time to first working system | Weeks | Six to twelve months (hire, ramp, build) |
| Cost shape | Fixed fee per build, retainer to run | £70k–£120k salary per engineer, ongoing |
| Breadth of experience | Patterns from many businesses | Deep knowledge of yours, over time |
| Risk if the first project fails | One fixed fee | A hire you now have to redeploy |
| Best when | Few, high-value problems; no existing AI capability | A steady pipeline of AI work; product is the business |
The mistake on each side
Hiring first. A single AI engineer without a defined problem list spends the first year exploring. The seat is justified by a pipeline, and most businesses don't know their pipeline until someone has audited the operation.
Renting forever. A consultancy that never hands over creates dependency. Insist on ownership: the code, the documentation, a named internal owner. Good consultancies design for their own exit.
The sequence that works
- Audit the operation to find and price the automation opportunities. This is the pipeline your future hire would need anyway.
- Build the first one or two systems with a consultancy on fixed fees. Real systems in production, owned by you.
- Decide on in-house with evidence: if the ranked register still has years of work on it, hire; if not, keep a run-and-improve retainer and spend the salary elsewhere.
M22's engagements follow this shape by design: audit, build, run, with everything owned by the client. Where a client wants to bring the work in-house, the handover is part of the deal, not a negotiation.
How much does an in-house AI engineer cost in the UK?
Roughly £70,000 to £120,000 in salary for a capable engineer in 2026, more in London and for senior profiles, plus recruitment, tooling and a ramp period of several months before a first system ships.
Can a consultancy hand over to an in-house team later?
Yes, if it was built that way: code and documentation owned by you, a named internal owner from day one, and a run-and-improve period that doubles as knowledge transfer. Ask for this before signing.
Written and maintained by M22, a London-based AI consultancy. Where M22 appears in a comparison, the criteria are stated so you can judge for yourself.
Want this answered for your business specifically? A thirty-minute call with M22 gets you an honest read on where automation would pay, and where it wouldn’t.
Book a call ↗