How to present the state of AI to the board of directors
The AI presentations reaching boards of directors belong, with few exceptions, to two genres. The first is the technology tour: twenty slides on what generative models are, a demo, a list of ongoing initiatives, applause and no decision. The second is the anxiety genre: a review of risks, AI Act ...
The AI presentations reaching boards of directors belong, with few exceptions, to two genres. The first is the technology tour: twenty slides on what generative models are, a demo, a list of ongoing initiatives, applause and no decision. The second is the anxiety genre: a review of risks, AI Act sanctions, news stories, with the effect of freezing every initiative. Both fail for the same reason, they do not give the board what a board lives on: comparable numbers, options with trade-offs, a precise request to deliberate upon.
The topic is less niche than it sounds, because the audience of boards that will have to face it is enormous: according to the Artificial Intelligence Observatory of Politecnico di Milano (2025 research, February 2026), 71% of large companies run active AI projects and only 9% have structured management. In the distance between those two numbers sit thousands of boards that sooner or later will ask "where do we stand?", and someone will have to answer in twenty minutes.
The opening number
The presentation that works opens with a measure, never with context. A maturity rating on a known scale (in our model, 0-5 with classes from A to D) gives the board in thirty seconds what reviews of initiatives fail to give in an hour: a position. "We are a class C, with Readiness at 3.1, Delivery at 2.8, Risk at 1.9 and Confidence at 2.5" is a sentence every director can interrogate, compare with next quarter, use to hold management accountable.
The number works under two conditions. The first is transparency about the method: where it comes from, on what evidence, with what repeatability, because an experienced board rightly distrusts self-produced scores. The second is honesty about the uncomfortable part, and here critical gates help: saying that the class stays capped until the systems register required by the regulatory framework exists conveys more credibility than ten slides of successes.
The three blocks that follow
After the position, the effective presentation stands on three blocks, each with its own deliberative function.
The first block translates gaps into exposure: what the company concretely risks by staying where it is, in terms the board already handles. The regulatory gap becomes exposure to sanctions and exclusion from tenders demanding governance evidence; the data gap becomes capital invested in pilots with a high probability of failure; the skills gap becomes dependence on suppliers. Each exposure should be quantified where possible and dated always, because the cost of inaction has a time profile and the board must see it.
The second block is the roadmap with its trade-offs, presented as a choice and never as a single plan to ratify. Two or three scenarios with different investment intensities, each with what it closes, what it postpones and what it risks, give the board the material to practise its trade, which is allocating capital across alternatives. The single plan asks for trust, the scenarios ask for a decision, and boards decide more willingly than they trust.
The third block is the ask, one and precise: approve the phase-one budget, appoint the owner with a mandate, put the twelve-month re-assessment on the agenda. Presentations without an ask produce the most frequent board minute in existence, "the board takes note", which is the formal way of not deciding.
The mistakes that lose the room
Three practices sink even well-built presentations. The first is jargon: talking about RAG, fine-tuning and agents to an audience that allocates capital moves the conversation from the board's terrain to the presenter's, and a board that does not understand postpones. Translating into business language is not a simplification, it is the job.
The second is the live demo, which steals minutes from deliberation and shifts attention from the state of the organisation to the capabilities of the tool, which was not the question. The third is presenting only successes: a board that discovers problems from other sources, an audit, an article, a whistleblower, stops using management presentations as a source, and that trust is not rebuilt in one meeting.
From event to routine
The single presentation, however successful, solves little if it stays an event. The underlying goal is bringing AI into the board's ordinary reporting cycle, with the rating on the dashboard next to financial KPIs, updated at every measurement, and a defined cadence for exceptions. It is the criterion behind the Executive Presentation we build inside the AI Rating, the deliverable designed for the board: strategic summary, positioning, recommendations, in a format a board can absorb and reuse.
If you have a board meeting scheduled in the coming months with AI on the agenda, preparing it from a real measurement changes the quality of the discussion: calendly.com/fabiolalli/zerofive, or hello@zerofive.ai. And if you have already given your last AI presentation, the retrospective test is merciless but useful: what did the board actually resolve?