Talking to your board or investors about AI

The short answer

A board or an investor group needs three things on AI, and almost every question they ask is a version of one of them. Where is it being used in this business, by whom, and on what data. What has it returned, measured against a baseline, and what has it cost. And what could go wrong, with what mitigations. Presenting those three in one page each quarter prevents both failure modes: the hype deck that promises transformation and creates expectations nobody can meet, and the silence that suggests management is not paying attention to something the board reads about constantly. Numbers carry the discussion better than narrative, so bring hours saved, cost per task, error rates and cycle times with the baseline and the method stated, including the ones that did not move. And have the risk answers prepared, because the questions about data, dependency, compliance and vendor behaviour will come, and hesitating on them is what turns a routine update into an agenda item.

The one-page update

Section	What it contains	Length
Where AI is used	The register: process, owner, data touched, whether it affects people	A short table
Results	Measure, baseline, current, change, for each use, including the flat ones	A short table
Spend	Subscriptions, usage, project cost, against budget	Two lines
Risks	Top three with mitigation and owner; any incidents since last time	Three lines
Governance	AI Act classification where relevant, data decisions, policy changes	Two lines
Next quarter	What is planned, why that, expected measure	Three lines
Deliberately not doing	What you have decided to skip and why	Two lines

Preparing the conversation

When the board pushes for more ambition

Sometimes the pressure is to do more, faster, because a competitor announced something. The useful response is the sequence: here is what we automated, here is what it returned, here is the data and process work that makes the ambitious thing possible, and here is when it becomes sensible. Boards respond well to a plan with evidence behind it. If the ambition is genuinely warranted and under-resourced, say what it would take in money and people, and let the board decide with real numbers rather than pressing for enthusiasm.

What this means for you

Give your board one page each quarter: where AI is used, what it returned against baselines, what it costs, the top risks and incidents, and what is next and deliberately not next. Measure before you start, report the misses, prepare two-sentence answers to the standard risk questions, and escalate governance items early. It prevents both hype and silence, and it makes the AI conversation a routine item rather than a recurring source of anxiety.