Your board does not need a prompting course

When a company decides its leadership should “get up to speed on AI”, the default is a version of what everybody else gets: a tour of tools, some prompting tips, a few impressive demonstrations. The executives leave able to write a better email with an assistant and no better prepared for the part of the job only they can do.

That part is deciding. Which proposals to fund, how much to spend, how much risk to accept and how to tell afterwards whether it worked. An executive session should be built around those decisions, not around the tools.

What executives actually need to be able to do

Tell a real proposal from a fashionable one. A real proposal names a decision or a process, an owner, a cost to build and to run, and a way to tell afterwards whether it worked. A fashionable one names a technology and a competitor who is supposedly using it.

Ask about the cost of running, not only of building. Running costs — usage, monitoring, the people who check the output, the work whenever the model provider changes something — are where business cases most often go quiet. One question in the board meeting changes that.

Understand how these systems fail. They rarely stop working. They become gradually less right while continuing to look fine. An executive who knows this asks how quality will be monitored and who will notice when it slips; one who does not treats go-live as the finish line.

Know where the regulatory lines run for this company. The EU AI Act places obligations not only on those who build AI systems but also on organisations that use them, and what applies depends on what a system is used for. Executives do not need the text of the regulation. They need to know which of their plans fall into which category, and when to ask for a legal opinion.

Calibrate expectations on their own processes. What these systems do well — drafting, classifying, finding, summarising — and where they need a person checking. Not in general, but in your invoicing, your service desk, your quality reports.

Infographic

Where a proposal sits, and what it has to prove

Draft to editSuggestion for a decisionClassification, checkedAction after approvalUnattended action

Human checks allUnattended

What the half-day looks like

Half a day is enough if it is built on your material rather than on general examples. In practice that means the proposals already on the table, or the processes people keep mentioning, examined with the questions above. No tool training, and no demonstrations of other companies’ results.

The room should hold the people who approve AI spending and the people who will be asked to own the results — up to about a dozen. Any larger and it turns into a presentation. Unlike a hands-on workshop, a session like this works well remotely. And before it, a good provider asks for your proposals and processes, not for a list of topics you would like covered.

It should end with something concrete. A shortlist the leadership agrees on, a rejected idea with the reason written down, or a set of questions every future AI proposal has to answer before it reaches the board. That last one is often the most durable output, because it keeps working long after the session itself is forgotten.

Why a shared vocabulary matters more than it sounds

In many companies the executive, the manager and the engineer use the same words for different things. “Agent”, “accuracy”, “pilot” and “done” mean one thing in the board pack and another in the team. Arguments that look like disagreements about strategy are often disagreements about definitions.

A session that settles a handful of those definitions, and leaves them written down, saves a good deal of time later — mostly in meetings that would otherwise end without anyone noticing they had been arguing about different things.

What it costs, and the case for it

The cost is half a day of the whole leadership team, and that is not small. The case for it is not that executives ought to understand AI. It is that they approve the spending and accept the risk, and without the right questions a weak proposal looks exactly like a strong one. One avoided bad project pays for a great many sessions like this.

That is how the executive session in our AI workshops and training is built: half a day, on your own proposals and processes, ending with a decision or a written set of questions, and with no tools on the agenda.