How to Present AI Strategy to a Board: Lead With the Decision

Key Takeaways

→Lead the AI item with what you want decided, what it costs and when you'll report back. Directors can test a decision; a tutorial gives them nothing to test.
→Deloitte's 2025 survey found two-thirds saying their boards had limited or no AI knowledge, and AI reaches every meeting at a minority of boards, so expect a short slot, and not many of them.
→Five slides carry the evidence: the recommendation, readiness judged by its weakest area, dated exposure, the ask in money, people and 90 days, and the quarterly measures.
→Price governance as a percentage of the AI budget. The 11x multiple this piece once used was wrong, and two published studies make 10% a defensible starting share.
→Answer "this slows us down" with the median review time, reported every quarter on the board's own dashboard.

An AI item at a board meeting should open with the decision you want, what it costs and when you will report back, all inside the first minute. Directors who say their boards know little about AI need a decision they can weigh. Put a tutorial first and it eats the time they'd have spent weighing it.

The usual instinct runs the other way. The board is new to AI, so the presenter explains AI first and leaves the ask for the end, when time is short. I think that gets the order backwards. This piece is for the CIO or CTO who has been handed the slot, since that's who most boards talk to about AI.

Why Lead With the Decision When the Board Knows Little About AI?

Because the knowledge gap is real and the time is short. Deloitte's 2025 board survey had 695 respondents, 84% of them board members. Two-thirds said their boards still had "limited to no knowledge or experience" with AI. Nearly a third said AI was not on the board agenda at all. Only 17% said their boards take it up at every meeting. Protiviti and BoardProspects, surveying 772 board members and executives at the end of 2025, put the every-meeting figure at 26%. In the same Deloitte survey, 72% said their boards talk to the CIO and CTO about AI.

Directors don't need to know how a model is trained to judge an investment. They judge capital requests every quarter by asking what it costs, what could go wrong, who is accountable and how they will know. Give them a decision and they can start testing it from the first slide. Hand them a primer instead and the item ends with a board that feels better informed and has nothing on the table.

The education still has to happen, just in its own session. Deloitte found that roughly half of organizations give their boards foundational AI education, and that 59% of individual board members have gone looking for it themselves. NACD asked 84 private-company directors in 2025 to name the single most important factor in overseeing AI: 27% chose director education, 25% a management team with the right AI focus and expertise, and 17% the information and metrics the board receives. The last two are yours to supply. Ask the chair for a separate teaching session, and send a one-page glossary with the papers.

Five Slides, With the Recommendation on the First

Each slide answers one question a director will ask whether or not they say it aloud. A slide that answers none of the five comes out of the deck.

Slide	The question it answers	What goes on it
1. The recommendation	What are you asking us to decide?	The decision, the amount, the date you report back, and the one result it buys
2. Readiness	Can we do this?	Scores by area, with the weakest named as the limit
3. Exposure	What could go wrong?	Top three risks, the rules with dates, a cost range
4. The ask	What exactly do you need?	Money, named roles, and three dated deliverables
5. Reporting	How will we know?	The six quarterly measures, one owner, the next review date

Slide 1, the recommendation. Say it in one sentence a director could repeat at dinner: "We ask the board to approve $X for an AI governance function and AI as a standing item each quarter, with the first report at the March meeting." Add the single business result the money buys. Everything after this slide is evidence for it. If your chair takes papers as read, make this sentence the paper's first line, and the first minute happens before the meeting.

Slide 2, readiness. Show where the organization stands and name the weakest area, because readiness is set by the weakest of the five pillars (strategy, data, talent, process and governance), not by their average. "Our data is a 1 out of 5, and it caps everything else" tells a director more than an overall 3.4.

Slide 3, exposure. Three risks, each tied to a system you actually run. Put dates on the rules that apply. In the EU, high-risk uses listed in Annex III of the AI Act, such as hiring and credit scoring, carry their obligations from 2 December 2027 under the Digital Omnibus, Regulation (EU) 2026/1744. Give the cost as a range with its assumptions written beside it. EY surveyed 975 C-suite leaders at companies with more than $1 billion in revenue in 2025: 99% reported financial losses from AI-related risks, and 64% lost more than $1 million. The Liability Ledger counts each obligation a live system ships without as a debt, kept in five accounts, at a rate that rises the longer it stays unpaid.

Slide 4, the ask in detail. Money, named roles and three deliverables with dates. Minimum Viable Governance gives every AI system three things in 90 days: a place on a list, a record of how it could fail, and a named person who can pause it. Those make a good set of day-90 promises, because a director can check each one.

Slide 5, reporting. What the board gets every quarter: the measures, the executive accountable for them and the date of the next review. This slide turns a one-off approval into oversight that keeps running. The six measures, with the thresholds behind each colour, are in A Board AI Dashboard That Is Always Green Is Broken.

How Do You Ask for Governance Money Without Hearing "This Slows Us Down"?

Start by not promising a multiple. An earlier version of this article told presenters to claim an 11x return on one governance hire, multiplying EY's average loss by an incident rate from a different McKinsey survey. That sum was wrong, and the companion piece on the ROI of AI governance explains why. A CFO on your board can take a multiple like that apart in an afternoon.

Ask for a fixed slice of the AI budget instead, topped up for each high-risk system on slide 3, since a cheap bought tool can carry the largest exposure. I found two published reference points. In IBM's 2025 survey of 915 executives, the top quartile spent 10% or more of its AI budget on AI ethics and reported higher operating profit from AI than the bottom quartile, which spent 5% or less. PwC's 2025 model compared companies spending only what compliance requires with companies spending an extra 10% of their AI budget on responsible AI, and found the second group ahead on value and revenue, results PwC calls "directional rather than exact". Neither proves governance caused the gains, and the two measure different things: IBM's 10% is where its top quartile starts, while PwC's sits on top of compliance spend. Still, both make a tenth a number you can defend without promising anything.

Then take the slowdown objection seriously, because it is right about governance that has no clock. A review queue with no deadline does slow teams down. So promise a turnaround, for example a decision on a low-risk use within a set number of working days, and put the median review time on the quarterly report. If governance does slow the business, the board will see it in your own numbers, and an annual check by internal audit or the risk function means the board isn't relying on the presenter's word. Directors trust a claim they can check against their own dashboard far more than a speed figure from someone else's survey.

What Should the Board See Every Quarter?

Six measures cover risk, performance, compliance and value. Investors are reading board oversight too: in Glass Lewis's review of 2025 proxy statements, 54% of S&P 100 companies disclosed board-level oversight of AI, and only 28% disclosed both oversight and an AI policy.

AI systems in use, by risk tier, with the share that have a current risk assessment and a named owner who can pause them.
Systems outside their agreed limits (accuracy, drift, override rate), and how many breached them this quarter.
Serious incidents, near misses and complaints or appeals from the people the systems decide about, with the days from detection to the board.
Governance review: median days from request to decision, how many reviews required changes or a refusal, and any system that went live without a review.
Each dated legal obligation that applies, with its status and owner.
Value delivered against the approved business case for each material initiative, with the scale-up or stop decisions due.

Each red, amber or green against thresholds the board agreed before launch; internal audit or the risk function checks the numbers once a year.

Send the dashboard with the board papers, so the meeting is spent on the reds and ambers. The board AI briefing template lays out the five slides and these six measures on one printable page.

What If the Board Has Three People?

A startup board of two founders and an investor doesn't need slides. It needs one page, sent two days ahead, with the same order: the decision and amount in the first paragraph, readiness and exposure in the second, the deliverables and the measures you will report in the third. Open the meeting by pointing to the first paragraph.

The strongest objection to all of this is that a board cannot approve what it does not understand, and leading with the ask will feel like a hard sell to directors who are still learning. I take that seriously. But the alternative, explaining first and asking last, risks a deferral, because the ask arrives when nobody has time left to test it. A board that understands the decision it is being asked to make, with its cost and its exposure, understands enough to approve it or to say what it needs to see first.

What to Do Before Your Next Slot

Write the first sentence of slide 1 and read it to someone outside the AI team. If they can't repeat it back, rewrite it. Then score readiness and name the weakest area, list the three risks with their dates, cost the ask as a share of AI spend, and fill in the six measures with this quarter's numbers, even where a number is zero or unknown. An "unknown" in the incident column is itself a finding the board should hear.

Sources and Changes

Deloitte Global Boardroom Program, "Governance of AI: A critical imperative for today's boards", 2nd edition, 2025. 695 respondents in 56 countries, January to February 2025.
Protiviti and BoardProspects, "Only 26% of Directors Discuss AI at Every Board Meeting", press release, 18 March 2026. 772 respondents, fourth quarter of 2025.
NACD, "2025 Private Company Board Practices Oversight Survey Data Pack: Artificial Intelligence", 26 August 2025.
Sarah Wenger (Glass Lewis), "US AI Oversight Through Three Lenses", Harvard Law School Forum on Corporate Governance, 11 March 2026.
EY, "EY survey: companies advancing responsible AI governance linked to better business outcomes", 8 October 2025.
IBM Institute for Business Value, "The AI ethics trust engine", October 2025.
PwC, "Quantifying the value of Responsible AI", 7 August 2025 (archived copy).
Regulation (EU) 2026/1744 (the Digital Omnibus on AI), Official Journal, 24 July 2026, new Article 113, third paragraph, point (c).

All sources checked on 6 October 2026.

Changes on 6 October 2026. Cut from about 3,900 words to an Argument under a new headline, with the recommendation moved to the first slide. Withdrawn: the 11x return on a governance hire, which multiplied figures from two different surveys, and speed and incident figures from a vendor blog that names no study. Removed: a claim credited to Deloitte about presentations that overstate readiness, which is not in the Deloitte report; a share of boards that receive AI metrics, which I could not trace to its underlying survey; a presentation timing rule from a practitioner blog; and six custom graphics, a download box and a canvas prompt. The quarterly measures now match the board dashboard piece.