Should I hire someone in-house for AI or use an outside company?

Job ads are full of AI skills now, so it is a reasonable thing to wonder: if everyone else is hiring for this, am I falling behind by not having one of those people on payroll?

For a business your size, almost certainly not. Here is the maths and the reasoning.

What does an in-house AI hire really cost?

Start with salary. The median advertised salary for an AI engineer in the UK is £88,750, based on vacancies posted in the six months to June 2026. Add employer National Insurance, pension, kit and recruitment fees and you are past £100,000 a year before they have automated a single thing.

The salary is not even the real problem. The real problem is workload shape. A small business does not have a year of AI work; it has an intense build phase of a few months, then refinement, then upkeep. The build phase needs a senior person. The upkeep does not. So you end up paying a senior salary for what becomes a part-time job, and good engineers in that position get bored and leave, taking everything they know about your systems with them.

Compare that with buying the build: a bounded project with a support arrangement afterwards, at a fraction of one year's salary. We have set out what those projects tend to cost in our guide to AI automation costs for UK businesses. Below about 50 staff, the hire very rarely pencils.

Half the AI job ads I read are companies hiring for a problem they have not defined yet. Do not join them.

When does hiring in-house make sense?

Three situations, honestly:

• AI is the product. If what you sell is software and the AI is your edge, you need that capability inside the building. No argument.
• You have a genuine pipeline of work. Not one automation project, but a continuous stream that will keep a builder busy for years.
• You are big enough to feed them. At 100+ staff, enough processes exist that a dedicated person stays useful. At 15 staff, they will run out of road by spring.

If none of those describe you, the in-house hire is a £100k bet on having enough work to justify it. Most small businesses lose that bet.

What should I watch out for with an outside firm?

The honest counterweight: outside help has its own traps.

The big one is dependency. Some agencies build things only they can maintain, on accounts only they control. Insist on owning the code, the accounts and the data from day one, and get the handover documented. If a firm resists that, walk away.

The other is the strategy-first sell: a consultancy that wants to run a discovery phase and produce a roadmap before anything gets built. You do not need a roadmap. You need one working system, then another.

And whichever route you take, do not hire a head of AI before you know what you want built. That is a strategy salary with nothing to direct.

I have made the fuller, more opinionated version of this argument in stop hiring for AI skills you should be buying as a system. The short version stands: buy the system first, hire when the workload proves itself.