Quick answer: A board report on AI should fit on one page and answer five questions: how much are we spending, what has it returned, which programs are working, what did we decide, and what are the risks. Use the same metrics every quarter, lead with dollars, and show realized value separately from potential value.
Key takeaways
- Boards are asking about AI return because the spend is now material. Worldwide AI spending is forecast at about $2.75 trillion in 2026 (Gartner, November 2026).
- Consistency builds credibility. The same five sections every quarter let directors see trends, not just snapshots.
- Decisions matter as much as results. Show what was scaled, changed and stopped.
What should an AI board report include?
- Spend summary. Total fully loaded AI spend this quarter and year to date, against budget, split by program stage.
- Return summary. Realized value to date, capacity value to date and portfolio payback.
- Program scorecard. The top five to ten programs by spend, each with cost, outcome change, realized value and status.
- Decisions. Programs scaled, changed or stopped this quarter, with one line on why.
- Risks and next quarter. Material AI risks, governance updates and the decisions planned for next quarter.
Section 1: Spend summary
| This quarter | Year to date | Budget YTD | Variance | |
|---|---|---|---|---|
| Experiments | $[ ] | $[ ] | $[ ] | [ ]% |
| Pilots | $[ ] | $[ ] | $[ ] | [ ]% |
| Scaled programs | $[ ] | $[ ] | $[ ] | [ ]% |
| Unallocated AI spend | $[ ] | $[ ] | ||
| Total, fully loaded | $[ ] | $[ ] | $[ ] | [ ]% |
Fully loaded means licenses, model usage, infrastructure, implementation, internal people and governance costs.
Section 2: Return summary
| Measure | Year to date | Definition |
|---|---|---|
| Realized value | $[ ] | Savings or revenue that has changed the P&L |
| Capacity value | $[ ] | Time or capacity freed but not yet removed or redeployed |
| Programs with measured outcomes | [ ] of [ ] | Programs with a baseline and a measured change |
| Portfolio payback | [ ] months | Cumulative spend divided by monthly realized value |
Section 3: Program scorecard (example)
The figures below are illustrative.
| Program | Stage | Annual cost | Outcome metric | Change vs baseline | Realized value YTD | Status |
|---|---|---|---|---|---|---|
| Support deflection | Scale | $700K | Cost per resolved ticket | -24% | $702K | On plan |
| Invoice processing | Pilot | $180K | Cost per invoice | -11% | $40K | Watch |
| Sales proposal drafting | Pilot | $220K | Proposal cycle time | -30%, win rate flat | $0 | Change |
| Code assistant | Scale | $1.1M | Cost per shipped change | Not yet measured | $0 | Baseline needed |
| Contract review | Experiment | $60K | Review hours per contract | Too early | $0 | Review in Q2 |
Section 4: Decisions this quarter
- Scaled: [program], because [evidence].
- Changed: [program], from [old scope] to [new scope], because [evidence]. New review date: [review date].
- Stopped: [program], because [no measurable outcome after X months]. Budget of $[ ] returned to the pool.
Section 5: Risks and next quarter
- Material AI risks and incidents, and how they were handled.
- Governance changes, such as new policies, vendor reviews or regulatory updates.
- Decisions planned for next quarter and the evidence that will inform them.
Tips for presenting AI to a board
- Lead with dollars, then explain with activity data. Directors want cost and return first.
- Never report time saved as return without saying what happened to that time.
- Show stopped programs. It signals discipline, not failure.
- Keep definitions fixed. If a metric changes, say so and restate the prior quarter.
Frequently asked questions
How do I report AI ROI to the board?
Use one page with five sections: spend, return, a program scorecard, decisions taken and risks. Lead with fully loaded cost and realized value, and use the same metrics every quarter.
How often should the board get an AI update?
Quarterly is typical, aligned with the financial close. Material programs or incidents may need ad hoc updates.
Should the board see adoption metrics?
Only as supporting context. Adoption explains why a program is or is not delivering, but it is not a return.
Who should present the AI report?
The CFO, ideally with the CIO or CTO for technical and risk questions.