The hidden cost of your first AI hire

“We are hiring a Head of AI.” It is the most common sentence on the board agenda this year. It is also, in our experience, the slowest, riskiest, and most expensive route to actually being AI-ready.

This is not an argument against in-house AI talent. It is an argument against treating “find the right person” as the first step. We have watched a dozen companies lose six to twelve months to that decision over the last eighteen months. Here is what it usually looks like up close.

1. The fully-loaded cost is bigger than the offer letter

A senior AI leader in London, Berlin, or Amsterdam right now is asking £180k–£260k base, plus equity, plus benefits, plus the recruiter’s 25%–30% finder fee on year one. Add a sign-on, a relocation budget, the laptop, the conferences, the OpenAI bill they will run up while exploring, the ML platform subscription they will insist on, and the contractor they will hire because the first project is already late. Year-one total cost of a single hire lands between £350k and £500k for most of our clients.

For comparison: a full Cravings AI engagement — readiness audit, anchor build, production handover, three months of enablement — typically lands between £120k and £280k. With a written brief, fixed scope, fixed timeline, and a system in production at the end.

2. Time-to-hire eats the quarter

The market for senior applied-AI talent is the tightest we have seen in any engineering specialism in a decade. Four to six months from “we should hire” to “they start” is the median. Then another three to six months to ramp — to build relationships, learn your data, find the right first project, earn the buy-in to actually ship it.

That is nine to twelve months before the first measurable AI outcome lands in production. In a market where your competitors are shipping in nine to twelve weeks, that is a meaningful loss.

3. One person has seen N projects. A team has seen N times more.

The hardest part of AI work right now is not the model — it is pattern recognition. Knowing which evaluation approach actually catches the failure mode you care about. Knowing which orchestration framework will outlive the next quarter. Knowing when a problem is a RAG problem and when it is just a data-cleaning problem in a trench coat.

A great in-house hire arrives with three to six AI projects of personal experience. A team that does this for a living arrives with several dozen, across industries, at different stages of failure and success. That breadth is the thing you cannot buy with a single salary, no matter how high.

4. Single point of failure, single point of bias

One AI hire is one opinion about which model provider to bet on. One opinion about whether to build agents or fine-tune. One vacation, one illness, one competing offer, one paternity leave that takes the entire AI roadmap with them.

We have been the call after that exit happens. The Slack messages start “we have a problem — they left and the system is a black box.” That is an expensive call to need to make.

5. The AI lead does not actually want to be alone

Worth saying plainly: the great AI hires we know — the people you would actually want — do not want to be the only person in the building who has done this before. They want to inherit something that works, with a team that has tooling and tests and runbooks, so they can spend their energy on the next thing, not on assembling the basics from scratch. A solo “Director of AI” role is one of the least attractive offers on the market right now, which is precisely why filling it takes six months.

What we actually recommend

Use a partner to compress the first nine to twelve months into nine to twelve weeks. Get the readiness audit. Ship the anchor project. Build the evals, the runbooks, the observability, the documented architecture. Then hire the in-house lead — into a role that is now attractive, into a system that is now live, with a brief that is now written down. They will ramp in weeks instead of quarters, and the offer letter you write will be for less than what the same person would have demanded six months earlier when the job was “build it all alone.”

That is the path we walk most of our clients down. It costs less. It ships faster. It produces a better hire at the end. The only thing it does not do is satisfy the board’s instinct that the first action on an AI strategy must be a recruiter brief.