An AI business case template a UK SME board can approve in one meeting.

The published evidence describes what a weak case looks like. Gartner names unclear business value, escalating costs, poor data quality and inadequate risk controls as the four causes of abandonment. DSIT found 60 per cent of UK businesses cite limited AI skills as a barrier. The ONS found 62 per cent of firms citing an expertise gap are training their own staff rather than hiring help. Each of those is a section a board will ask about, and a case that has no answer for one of them is a case that gets deferred.

The other failure is length. A twelve-page case is read by nobody and approved by the person who wrote it. One page, with the numbers, is read by everyone and argued with, which is what approval looks like.

Replace the bracketed items. Keep the headings.

The baseline is the section that decides the case, and it cannot be estimated from memory. Count the units for two weeks. Ask everyone who touches one how long they spend, then watch a few to check. Record arrival-to-action from the timestamps you already have in the inbox or the system. Count the ones that had to be redone. Multiply hours by a loaded rate that includes employer costs, not the salary alone. Our 90-day ROI guide sets out the method and a worked example, and the readiness audit exists to do this measurement for firms that would rather it was done for them.

Use the published UK ranges and say so. Our own prices, which are fixed against a written scope, are £2,500 to £6,000 for an adoption audit credited against a build, £4,000 to £15,000 for a first implementation, £8,000 to £25,000 for a custom AI system and £8,000 to £30,000 per phase for an application. Published UK market ranges surveyed in our cost guides sit around those bands, with day-rate suppliers higher and less predictable. The running cost is the line most cases omit: at September 2026 prices a small model costs a fraction of a penny per typical document and a text message costs more than the model behind it, so budget the channels, not the tokens. Our running-cost guide gives the arithmetic.

Section five is the ICO’s expectations in four lines: the data, the provider and region, the lawful basis, and whether a DPIA is needed, which it is where the processing is likely to be high risk under the ICO’s screening criteria. A business plan of the AI tool, a data processing agreement and a written policy answer most of it. Section six is the four Gartner causes turned into controls: an acceptance threshold as a number, a cost ceiling that has been watched firing, a defined failure path, and ownership of the code and accounts so the firm is never dependent on the supplier. A board that sees those six lines has seen a firm that has done the thinking.