Perspective · AI Strategy
What boards should ask before approving an AI budget
Five questions that separate a working AI plan from a slide deck.
In short: Before approving an AI budget, a board should ask which decisions will change and who signs them off, what the cost per transaction is at full scale, what the main risks are and how they are controlled, who owns delivery, and how the board will know within six months whether it is working.
AI proposals reach boards with ambitious claims and limited detail. These five questions help directors test whether a plan is ready for investment.
1. Which decisions will change, and who signs off?
A credible plan names the business processes and decisions that will change, such as how claims are reviewed or how maintenance is scheduled, and the executive accountable for each. If the plan only describes technology, it is not yet a business plan.
2. What does it cost per transaction at full scale?
Pilots are cheap. Ask for the expected cost per task or transaction when the system is used by everyone who will use it, including software, usage charges, integration and support. Compare that with the current cost of the same work.
3. What are the risks, and how are they controlled?
- Data protection and confidentiality of the information the system uses.
- Errors and how they are caught before they reach customers.
- Dependence on a single vendor, and the exit plan.
- Regulatory requirements in the sectors and countries involved.
4. Who owns delivery and adoption?
Ask who will change processes, train staff and adjust targets. Budgets that fund licences but not adoption tend to produce tools nobody uses.
5. How will we know in six months if it is working?
Agree two or three measures, such as time per case, error rate or customer response time, with today's baseline. Set a review date and a decision: scale, adjust or stop.
Frequently asked questions
What should be in an AI business case?
The decisions and processes that will change, accountable owners, full-scale cost per transaction, risks and controls, adoption plans and measurable targets.
How should boards oversee AI?
By approving investments against clear measures, requiring a register of AI uses and risks, and reviewing results at fixed intervals.
How soon should an AI investment show results?
A well-scoped initiative should show measurable progress against its baseline within about six months of going live.