Insights
How do I build a board-ready business case for AI?
By giving the board a decision it can vote on. Most AI cases arrive as a tour of capabilities, all what the models can do and how good the demo looked, and leave the board where it started: interested, with nothing to approve. A board-ready case answers the four questions a board asks of any investment: what is it worth, what will it cost, what is the risk, and what are you asking us to fund.
What a board actually wants to see
A board is not evaluating AI. It is allocating capital under uncertainty, the same as it does for any proposal. It wants a clear line from spend to value, an honest view of what could go wrong, and a specific ask it can vote on. The technology is a means. Lead your case with it and you have answered a question the board did not ask.
So lead with the value, keep the capability for the appendix, and frame the whole thing as a choice the board is being asked to make.
Start from the value
The strongest cases begin with a problem the business already wants solved and a number already on it: cost-to-serve, cycle time, error rates, churn, hours lost to a manual process. AI is then presented as the way to move that specific number, with a credible estimate of by how much. "We could use AI here" is a hope. "This process costs us £X a year and here is how we take a third out of it" is a case.
If you cannot attach AI to a number the board already cares about, you do not have a case yet. You have a research request, and it is worth naming as one.
The four things the case must contain
Everything a board needs fits in four parts.
| Part | The question it answers | What good looks like |
|---|---|---|
| Value | What is this worth, and how confident are you? | A specific number tied to a known problem, with a range, not a single hero figure |
| Cost | What does it really cost to get there? | Tools plus the change and adoption work, the part most cases forget |
| Risk | What could go wrong, and how is it managed? | Data, governance, EU AI Act exposure, and the risk of doing nothing |
| The ask | What are you approving, and what does success look like? | A specific decision, a timeframe, and the measure you'll be judged on |
The two parts most cases underweight are cost and risk. Under-counting the cost by leaving adoption out is how projects run over and confidence erodes. Ignoring the downside, including the governance and regulatory exposure, is how a board that has read the headlines quietly loses faith in the proposer.
Size it honestly
The fastest way to lose a board is an ROI number that looks too good. Inflated returns get discounted on sight by people who have seen a hundred business cases, and they take the proposer's credibility with them. A defensible range, with the assumptions visible and a conservative case included, beats a spectacular point estimate every time. Boards fund proposers they trust more readily than projects that dazzle.
Include the cost of adoption alongside the cost of the tool. And include the cost of doing nothing, the standing-still position against your peers and your own targets, because that is the alternative the board is really choosing between.
Make it a decision the board can vote on
End with the ask, sharply. What are you asking the board to approve, for how much, over what period, and how will everyone know in six months whether it worked. A pitch invites discussion and tends to drift, whereas a decision invites a vote. The difference between a case that stalls and one that gets funded is usually that the second told the board exactly what it was being asked to decide.
Get value, cost, risk, and the ask onto a page in the board's own language, and AI stops being a topic the board is curious about and becomes a decision it can make.
Firestarter's six-week accelerator produces exactly this: a board-ready roadmap with the value quantified, the cost and risk counted honestly, and an ask the board can vote on.