How to write an AI business case.
Moonlabs is the operator-led AI Academy in Derby. We run three live companies — Homemove, home.co.uk and homedata.co.uk — and we teach senior operators to ship AI internally, defend the spend, and walk a board through it. Three pillars: Coding, Commercials, Investment. Twelve weeks. £6,000 — usually Apprenticeship-Levy fundable.
Moonlabs is what we are. Two operators — James Freestone and Louis O’Connell-Bristow — who have written, defended, approved and rejected dozens of internal AI business cases across Homemove, home.co.uk and homedata.co.uk. We have sat on both sides of the table — the one writing it and the one cutting it. The structure on this page is the one our own CFO-equivalents actually approve.
The Academy is what we do. A twelve-week, in-person, twelve-student cohort in Derby. You build a real AI tool inside your own workflow. You sign your team off as the first paying user. You write a deck and a financial model that survive a board read. Coding, Commercials, Investment — the three pillars taught in equal weight every week, with this page’s deliverable being the artefact your board approves.
Why this page exists. Most AI business cases are bad in a specific way: they lead with capability (“we will use GPT-5 to…”) when boards care about outcomes (“we will cut reconciliation time by 40% in two quarters”). They quote vendor pricing instead of unit economics, skip the failure case, and forget that boards in 2026 are sceptical because they have seen too many badly-built cases land. You leave the Academy with a case your board approves and a deployed tool that proves it — before twelve months goes by.
Coding · the deployed tool that proves the case
The strongest business case is one where the pilot is already running. By week six the AI tool is live in your workflow with a real eval suite catching regressions. The cost, the time-saved, the failure-mode behaviour — all measured in production, not modelled. A deployed internal AI tool by week twelve — the proof your board cannot argue with.
Commercials · outcome-first, unit-economics-honest
Lead with the number, explain the AI in the second paragraph. Real per-query cost, eval cost, observability cost, the human-in-the-loop cost. Baseline measurement. Named failure case and off-ramp. A 90-day pilot framing — boards approve £20k pilots; they sit on £2m programmes for six months. A business case under three pages by week ten.
Investment · approval, plus the next round of pilots
Once the first case clears, the second is half the work. Some Academy graduates go on to head AI strategy inside their employer, sponsor more seats on the next cohort, or spin the internal product out as a vendor (with the employer’s blessing). Sponsorship routes covered: Apprenticeship Levy, L&D budgets, R&D Tax Credit framing. Funding routes here.
Common questions.
What should an AI business case actually contain?
A one-page summary of the workflow being changed and the outcome you are targeting. A baseline measurement. A specific 90-day pilot, costed end to end (model, evals, observability, human work). The named failure case and off-ramp. A 12-month roll-out plan IF the pilot hits the target. That is the whole document; everything else is decoration.
How should I cost AI in 2026 — per seat, per query, or against headcount?
Against the workflow it changes. "Will save 1.2 FTE-weeks per month against a £450/month subscription" is the conversation boards understand. Vendor pricing models are a footnote, not the headline. More on pricing here.
How long should the business case be?
Three pages or fewer for the board version; a longer technical appendix for the InfoSec / risk team if asked. The biggest mistake is over-length — a 30-slide deck signals that you have not finished thinking.
Can my team fund this course on the case for building AI internally?
Yes — that is one of the most common funding stories. L&D budgets, AI upskilling allowances, the Apprenticeship Levy all work. Full funding routes here.
How does this compare to writing a fundraising pitch?
Same skill, different audience. Boards approve cost; investors approve growth. Both want specific numbers and a named failure case. The pitching page covers the founder version.
More Academy entry points.
The Academy is one course with many doors. Each of these pages is a different entry point into the same twelve weeks.
How To Validate An AI Startup Idea
How To Find Paying Pilots For AI Startup
How To Become An AI Engineer
How To Incorporate An AI Startup UK
How To Build An AI Moat
How To Launch An AI Product
Build it. Sell it. Defend it. In twelve weeks.
Tell us the workflow you would build a case around and we will walk you through it. James and Louis read every application personally and reply inside the week.
© 2026 Moonlabs Incubator. All rights reserved.