Why a fintech chose a Cravings pod over a Head of AI — then hired both

The brief: A UK fintech with about 180 employees needed to deliver an “AI-powered onboarding” capability in the next two quarters. The COO had built a spreadsheet comparing two paths: hire a Head of AI to run the project in-house, or engage Cravings as a partner. They asked us to help them stress-test the spreadsheet honestly — including the case against us.

The numbers, side by side

• Path A — in-house hire. Senior AI lead at £200k base. 28% recruiter fee. Sign-on. Equity. Loaded year-one cost: ~£420k. Estimated time-to-first-production: nine to twelve months, of which the first four were a search.
• Path B — Cravings engagement. Two-week readiness audit (£40k) plus a twelve-week anchor build (£190k) plus three months of enablement (£60k). All-in: £290k. Estimated time-to-first-production: thirteen to fifteen weeks from kickoff.

The COO’s honest pushback: “Year two looks worse for you. The Head of AI is still on the team in month thirteen. You are not.”

Our honest answer: yes, and we say so on the proposal. The right answer is to do both, in the right order. Use Cravings to ship year one. Hire the Head of AI into the working system once it exists. Their offer letter is smaller because the role is now attractive (a live system, a team, runbooks). Their ramp is weeks instead of quarters. And the brief you write for them is a real brief, because we wrote it with you while we were on the project.

What they decided

The COO took the proposal to the board with both options modelled to thirty-six months. The combined “Cravings then hire” path was £180k cheaper over three years, shipped seven months earlier, and produced a measurably better in-house hire when they finally ran the search at month eight.

What we shipped

• A KYC document-extraction agent that handled 73% of incoming applications without human review.
• A risk-flagging agent that ran in shadow mode for four weeks before being given write access to the case management system.
• An eval suite of 600 historical applications, regraded by the compliance team, with a pass-rate gate on every deploy.
• Documented architecture, deployment pipeline, runbook, and a thirty-page handover brief.

What changed

• Onboarding completion rate: 39% → 64% on the cohort the agent handled.
• Median time-to-first-deposit: 14 minutes → 5.
• Compliance findings on the post-launch audit: zero.
• Time from kickoff to live agent: 14 weeks.

The hire that came after

The fintech opened the Head of AI role at month seven, with a written brief that read “extend and own this working system” rather than “build AI for us from scratch.” They received twice the applications they had expected. The hire came in at £165k base — £35k below their original budget — because the role was now attractive to a stronger, more selective candidate pool. The Cravings pod rolled off two weeks after that hire started.

The COO’s note to us at handover, with permission to share: “I genuinely thought we were going to lose the year-one comparison and only win on speed. The board did not. They picked the version where the in-house hire walks into something rather than nothing. We will do this exact sequence again on the next AI surface.”