Ten minutes on the agenda, late in the meeting, after the numbers. A non-technical board, two of them investors, and the founder who handed you this sitting at the far end having said very little for weeks. Somewhere in those ten minutes the question is going to arrive, in one form or another. What’s our AI strategy?
You know the honest answer. You are still building it. The assessment is done, the operating-model thinking is straight in your head, the sixty-day plan is written. None of it has earned a return yet. The instinct under that kind of pressure is to make the plan sound bigger than it is, to promise sweeping change fast and buy yourself the room. That is the move that costs you the mandate later. The stronger choice is to say the honest thing, well.
What does the board actually want to hear in the first update?
The board wants business impact in the format it uses for every other investment decision, not a technical status report. It needs to see that you have a method, that you have read the current state honestly, and that you are asking for one defensible next step. Boards forgive a modest plan. They do not forgive a confident claim that later turns out to be air.
That reframes the headline question before it is even asked. The honest answer to “what’s our AI strategy?” is that the business needs an operating framework for becoming AI-capable, not a single document you can wave in the air. Prolego’s research makes the point sharply, that executives win funding by talking about organisational capability rather than technology. So you carry the operating-model thinking into the room and explain that the firm will mature across a few dimensions in parallel, the capabilities, the data, the people, the governance and the tools, rather than buy one clever system and call it a strategy.
The credibility comes from how you hold the gap. You are not pretending the work is finished, and you are not apologising for it being early. You are showing the board a deliberate path with a clear first step on it. What a board will not back is a status update that lists activity and asks for nothing in return.
When is the honest current-state read the right move?
The honest current-state read is the right move in almost every first update, and certainly when there are no returns yet to point to. It is the more credible position because it is harder to fake. You tell the board where the firm genuinely is today, what has been tried, what the early signals are, and what you do not yet know. That candour earns you the room for the bounded ask that follows.
The structure that carries it is the five questions boards reliably bring to AI, drawn together in ScaledAgile’s work on what every leader should be able to answer. How does AI fit the core business model. How will returns be measured. What risks are being managed. What capabilities are being built. What competitive advantage is being created. Answer those five honestly and you have answered “what’s our strategy?” without ever needing the word. The five questions are the spine of the update and the order in which a board’s mind actually moves.
There is a second reason candour wins, and it sits with the founder at the end of the table. Boards are increasingly wary when a founder hands the AI mandate off entirely and goes quiet, a pattern Spencer Stuart’s research on AI-era leadership names directly. An honest read keeps the founder in the work rather than letting them drift to the edge of it. When you present the current state plainly and name where you need the founder’s judgement, you pull them back into the decision instead of carrying the whole exposure alone.
When does promising sweeping change cost you the room?
Promising sweeping change costs you the room the moment the board treats your words as a commitment and then watches the timeline slip. The flashy update lands well on the day. It also sets a clock running that you cannot beat, and the person who set it carries the blame. An inflated first update is the most expensive kind of credibility to spend, because you spend it before you have earned any.
The discipline that protects you is to have a reasonable basis for every claim and to name the limits of a small early sample out loud. This is the principle behind the regulatory scrutiny of overstated AI claims, framed for ordinary purposes rather than securities law. If a pilot has reduced handling time in a single team over a few weeks, say exactly that, and say that the sample is small and needs more validation before you would scale it. A board reads acknowledged limits as competence. It reads unqualified certainty about new technology as either naivety or spin.
The same discipline shapes what you commit to. Instead of blanket approval to continue, ask for a specific bounded next step, funding for a small number of projects measured on incremental progress, as Prolego puts it, think big but start small. That is a request the board can weigh and grant. A vague ask to keep going invites the board to add to your list, or to defer, and either way you leave the room with less than you came for.
How do you translate technical progress into board language?
You translate it by replacing the technical fact with the business outcome and the decision it supports. A board cannot act on “we’ve implemented AI for document handling”. It can act on what that produced, which way the number is moving and what you are asking it to approve next. Every point should carry three things, a value, a trend and a clear decision, the test Reco’s board-reporting work applies to security metrics.
Drop anything that fails that test. Model accuracy, processing volumes, the name of the platform, none of it belongs in front of a non-technical board unless it becomes business impact first. The translation work is yours to do before the meeting, not the board’s to do during it. Reframe each line until it reads as outcome, direction and decision, and the board has something it can interrogate and approve rather than nod through and forget.
A board update without a decision point in it is the most common way these presentations fail. The room hears a tidy summary, has nothing to do with it, and moves on. So build every section towards a specific request. Not “we’d like to keep investing in AI”, but a named next step with the resource it needs and the outcome it is meant to produce, in the same shape the board uses for any other capital request, current state, trend, investment, expected return.
What should you commit to before you stand up?
Commit to the honest read, the five-question spine, one bounded next step, and the timeline reset, and commit to nothing beyond what you can stand behind. Before you walk in, decide the single decision you are asking the board to take and make sure every part of the update builds towards it. If a slide does not serve that decision, it does not belong in ten minutes.
The timeline reset is the commitment people skip and regret. Meaningful returns commonly take twelve to twenty-four months, and a board that expects change in a quarter will read a normal pace as failure unless you have told them otherwise. The first update is the place to set that expectation honestly, on the record, while it costs you nothing. Reset it now and you are managing an agreed timeline. Leave it, and you are defending a slip you could have pre-empted.
What you do not commit to is a number you cannot yet support or a date you have not stress-tested. Hold the line between confidence and overclaim. Be certain about the method, the read and the next step, and honest that returns are not in yet. That combination is what a board trusts, and it carries you into the next meeting with the mandate intact. If you want a second pair of eyes on the update before it goes to the board, book a conversation.
