Most boards now get an AI update. Very few get an AI dashboard. The difference matters, because an update is written by whoever is presenting and a dashboard is a standing set of lines that stay the same whether the news is good or bad.
There is a specific reason boards need the second thing. In our September data, the seat that reports the healthiest picture of AI governance is the seat boards hear from most. Ask the CEO seat whether anyone owns AI purchasing and 2 percent say no single seat does. Ask finance and it is 15 percent. Ask the technology seat and it is 30 percent. The governance gap is least visible from exactly where the board sits.
That is not management hiding anything. It is a structural property of who gets to present. A dashboard is the cheapest available defence against it.
I have written separately about the ten AI questions directors should ask management. Questions work in the room. This is the thing that should arrive in the pack before the room, every quarter, in the same shape.
Why single-seat reporting fails
Boards are used to reading numbers that come from one function because that function owns them. Cash comes from finance. Pipeline comes from sales. AI does not sit in one function, so a single-seat AI report inherits that seat's blind spot.
Our data shows the effect plainly. On payback, 70 percent of CEO-seat respondents expect measurable AI return inside six months, against 42 percent of finance-seat respondents, on seat bases of 80 and 52. Both numbers are honest answers from people describing the same technology. If the board deck carries the first number, the board is being shown the most optimistic reading available, and the seat that will be asked to prove it has already told us it expects something slower.
The fix is not to distrust the presenter. It is to require the counter-seat number on the same page.
The seven lines
For each line: what it is, who reports it, and what to compare it against.
1. Named owner of AI purchasing. Reported by: the CEO or COO. Cross-check: ask the CFO and the CIO or CTO to confirm the same name in writing. Our Ownership Vacuum measure exists because these three answers frequently differ. If they differ in your company, the board has found something before it becomes an incident.
2. Signer of the last three AI purchases above the approval threshold. Reported by: finance. Cross-check: the purchase paperwork itself. This is the most revealing line on the dashboard and the one companies most often cannot fill in quickly. If nobody can produce three names in a week, approval is happening below the level the board assumes.
3. AI spend by funding source. Reported by: finance, split by where each pound came from rather than which vendor received it. Cross-check: total AI spend reconciled to the general ledger.
One category on that split deserves particular board attention: AI paid for with money earmarked for hiring. It runs at 21 percent across the finance lane and 34 percent at the CEO seat, on a base of 290. A company funding AI from hiring money is making a workforce decision, and it is often not being minuted as one. If the split shows a material share here, the board should expect a headcount plan to have changed somewhere, and should ask to see it.
4. Payback date on the largest three AI commitments. Reported by: the business owner. Cross-check: finance's own view of the same three. Where the two differ by more than a quarter, the board should ask why before the money is spent rather than after.
5. Proving-ROI blocker rate. Reported by: finance. In our data this rose from 53 percent of respondents in the cohort through July to 65 percent in the August cohort. A rising number is not necessarily bad news. It usually means the organisation has moved from deploying to being asked for proof, which is the healthier stage. A flat number near zero usually means nobody is being asked.
6. Agent inventory and access. Reported by: security. The count of AI agents running in production, what credentials each holds, and who owns each one. This is the line most likely to produce a surprise, because two thirds of the senior security leaders in our September rooms already rank agent access above every other AI concern, and boards rarely see the underlying count. Cross-check: ask whether the inventory is generated from systems or maintained by hand. A hand-maintained inventory is out of date the week it is written.
7. AI security funding status. Reported by: security. Whether AI security has a dedicated line, is funded case by case, is carved from the existing security budget, or has no spend. Only 37 percent of teams in our September data hold a dedicated line, on a base of 110, so the answer a board should expect is one of the other three. We track the shortfall as the Security Funding Gap. Cross-check: finance's view of the same budget, which sometimes disagrees.
What the dashboard is not
It is not a performance report on the AI programme, and it should not become one. Boards oversee; management operates. Every line above is designed to reveal whether a decision has an owner and a number, not to second-guess the decision.
It is also not a monthly artefact. Quarterly is right for most companies. The exception is the agent inventory, which should be available on request at any time, because the interval between an agent being granted access and a board hearing about it is the risk.
One number that predicts the rest
If a board can only add one line this year, make it line two: who signed the last three AI purchases above the threshold.
There is supporting evidence from the capital side. Investors who can name the AI owner in a portfolio company report nothing measurable in 9 percent of portfolios. Investors who say it is too early to name an owner report nothing measurable in 41 percent. That is a correlation drawn from investor perception on a base of 245, not a controlled study, and it does not prove that naming an owner causes returns. But it is consistent with everything the operator data shows: the companies that can say who decides are the companies that can say what happened.
Limitations
The board lane has no instrument of its own. Every figure above is read from the seats a board hears from, which is stated openly in the Board Director AI Governance Report and is worth holding in mind, because it means this dashboard is assembled from adjacent evidence rather than from directors themselves.
The seat cuts also differ sharply in weight. The CEO chair rests on 80 responses and finance on 52, both of which clear our publication floor. The technology chair rests on 23 and does not, so treat the 30 percent unowned figure as a direction the data points in rather than a measurement. All of it comes from executives applying to attend AI-focused sessions between March and August 2026, which is a group already engaged with the subject. Multi-select questions push several shares past 100 percent in total.
Where directors compare AI oversight
Open Future Forum runs a Public Board Member Dinner Series for directors of public and late-stage companies, alongside the CEO Executive Forum. If you want the underlying figures with their bases, the Board Director AI Governance Report sets them out, and membership is by application.
Last updated: September 9, 2026
Frequently Asked Questions
The Public Board Member Dinner Series
Open Future Forum convenes directors of public and late-stage companies for off-the-record dinners in Silicon Valley.