Why should AI be a recurring board topic?
In Deloitte’s 2025 survey, 66% of respondents described their board’s AI knowledge or experience as limited or non-existent, while 31% said AI was not on the board agenda. These are survey responses, not a measure of every board. They nevertheless point to the need for a regular discussion of AI-related value and risk.
For boards of private-equity-backed businesses, the investment horizon adds another consideration: AI benefits should be evidenced, and the technology should not create hidden operational or contractual exposures ahead of an exit.
Five questions for the next board meeting
1. Which AI initiatives support business priorities — and who is accountable for the outcome?
What management should explain: What business problem is each initiative addressing, what does success look like, and which executive is accountable? Benefits may take the form of EBITDA improvement, growth, service quality, faster turnaround or lower risk. Not every benefit will appear immediately in the P&L.
Evidence the board should see: A short portfolio of initiatives showing an accountable business owner, baseline, target, timing and how results will be validated by Finance or another relevant function.
Warning sign: The board sees numbers of tools, licences and users, but no evidence of operational or commercial outcomes.
2. What happened to our pilots — which have scaled, changed course or stopped?
What management should explain: What decision criteria were set before testing, when is each review due, and what evidence is needed before further investment is authorised?
Evidence the board should see: A pilot register with start dates, owners, measures, costs and explicit decisions: scale, revise or stop. Long-running pilots should have a documented rationale.
Warning sign: The pilot portfolio keeps growing without clear decisions. Six months can be a useful trigger for review, but it is not a universal cut-off.
3. Could we evidence the value of AI to a prospective buyer?
What management should explain: Which solutions are embedded in day-to-day processes, what has actually been measured, and where are we dependent on particular suppliers or individuals?
Evidence the board should see: A record of live deployments, credible before-and-after measures, ongoing costs, data-access rights, supplier contracts and business-continuity arrangements.
Warning sign: Forecast benefits are reported as realised savings, or a business-critical application depends on one individual.
4. Who owns AI risk and regulatory compliance?
What management should explain: Who maintains the inventory, assesses use cases, manages data protection and security, and escalates material incidents? The board oversees the effectiveness of these controls; it does not substitute for management or the risk function.
Evidence the board should see: An AI inventory, risk assessments and system classifications, acceptable-use standards, AI-literacy measures and regular reporting on incidents and remediation.
Warning sign: The business cannot say where AI is used, or assumes all AI Act obligations begin in 2027.
Regulatory position (November 2026): the EU AI Act applies in stages. Following Regulation (EU) 2026/1744, specified high-risk system obligations under Annex III apply from 2 December 2027 and those under Annex I from 2 August 2028. Other obligations, including those relating to AI literacy, may already apply. The precise requirements depend on the organisation’s role and the system in question.
5. Does the board understand the processes AI is supposed to change?
What management should explain: Which three or four processes most affect customers, cost and risk? What will change in how the work is done, and how have employees been prepared?
Evidence the board should see: A practical view of the process before and after, named process owners, a rollout plan and measures of sustained adoption.
Warning sign: Presentations focus on models, licences and demos, without explaining how work will actually change.
Turn oversight into a routine — without micromanaging
Make AI a standing item in the quarterly board review. Management can provide a one-page summary covering priority initiatives, benefits and costs, pilot decisions, material risks and matters requiring board attention. At least annually, review the wider AI inventory and the effectiveness of the governance arrangements. The cadence should reflect the company’s exposure and level of AI adoption.
The distinction matters: management chooses and delivers initiatives; the board challenges decision quality, accountability and the evidence of value. As MIT Sloan’s George Westerman argues, effective governance should help organisations progress safely rather than operate purely as a brake.
The takeaway
Five disciplined questions can move the AI conversation beyond technology presentations towards outcomes, ownership and risk. A board does not need to understand every model. It needs to know what the business has gained, what has been learned and what management proposes to do next.
A simple recurring board report and clear responsibilities are a good starting point for any company looking to strengthen AI oversight.
