The Business Case Template That Gets AI Projects Approved

Virtasant reported in September 2026 that only 14% of finance chiefs see a clear, measurable impact from their AI investments so far. Your finance director has read the pitches. Most of them promised savings and showed a product demo.

The same analysis says nearly half of generative AI projects are abandoned after the proof-of-concept stage. The usual cause is that nobody defined business value or success metrics at the start. A case with no baseline and no metric looks like one of those projects.

Do not promise fast payback without evidence. Virtasant cites Deloitte research showing that only 6% of organisations see AI returns in under a year. Some 37% wait two to three years. A realistic timeline protects your credibility. An optimistic one costs you the next request.

Pick one workflow. Contract review, report drafting, invoice queries or customer enquiries all work. Time it for two weeks before you deploy anything. Record the hours spent, the loaded hourly cost of the people doing it, the error rate and the rework. First AI Movers recommends the same two-week baseline in its April 2026 guide for European SMEs.

The baseline gives you a denominator. Without it you cannot show a saving. Your finance director also cannot check your claim a year later. If the number does not exist today, you cannot prove it changed.

Choose one lever. Cost reduction, revenue growth or risk reduction. Do not write "improved efficiency". State the pound figure. Here is an illustrative example. Four people spend five hours a week on one task. Their loaded cost is £30 an hour. That is £600 a week and £31,200 a year. Replace those figures with your own measured baseline.

Then state what share of that cost you expect to remove. Do not claim 100%. Say 30% or 50% and explain why. Frame it as the cost of producing an output, not as a headcount cut. Your staff will read this document too.

Cut any vendor-claimed improvement by 50% before you run a pilot. First AI Movers suggests this haircut for pre-pilot estimates. Run the pilot for about four weeks on the same workflow with the same metrics. Extrapolate the result over 12 months. Subtract the tool cost. Then reduce the total by 20% for adoption variance. Show every step of that calculation on the page.

Count the full cost. Include licences, training, integration, the time staff need to change how they work, and governance. Virtasant reports that cost underestimation for scaling generative AI often reaches 500% to 1,000%. If you leave costs out, your finance director will add them for you and trust you less.

Put a pound value on risk. Ask what a data breach, a wrong output or a regulatory fine would cost you. If you cannot give one figure, give a range and say how you built it. A risk you have priced is a risk you have thought about.

Book 30 minutes with your finance director this week. Ask for the loaded hourly rate for the roles in your chosen workflow. Ask which return they weigh most: cost, revenue or risk. Involve them in writing the case. Do not send it to them as a finished document.

Start the two-week baseline on Monday. Write a one-page case with four items: the baseline, the pilot plan, the risk-adjusted 12-month forecast and the break-even date. Then test it against three questions from the Virtasant analysis. What does the current state cost? What measurable financial impact will this deliver? What is the risk you remove worth in pounds? If you cannot answer all three, the case is not ready. Do not submit it.