An AI strategy isn't decreed in a slide deck: it gets decided. Notes from a day spent making an executive committee vote.
I spent a day locked in with the executive committee of a major financial player in Luxembourg. Not to present slides — to make them vote.
That day was the culmination of months of work: one-on-one interviews, from the CEO to field teams. A maturity audit. People trained before anyone decided anything. Workshops on culture, data, technology, governance. And dozens of use cases surfaced from the field — not invented in a meeting room.
The classic mistake: starting with the tools
Most AI strategies I see start from the wrong end: a list of tools, a POC, an 'AI project portfolio' living on the side. The outcome is known in advance: pilots that never reach production, teams watching the train go by, and a strategy that belongs to the vendor who wrote it.
Until the executive committee has made the calls itself, the strategy belongs to the vendor — not to the company.
What we did instead
- Train before deciding. You don't ask a board to arbitrate a matter it has never practised. The teams — and the executives — got their hands dirty before decision day.
- Make them vote, don't present. The company's AI ambition — its North Star — was put to a vote, amended, consolidated. The final wording is theirs, not ours. That's the whole difference between a strategy that is carried and one that is endured.
- Prioritise across three horizons — foundations, acceleration, transformation — and integrate everything into the global project portfolio. No theoretical AI portfolio that ignores real execution capacity, maintenance and ongoing projects.
The most useful moment wasn't a consensus
It was a disagreement. Two visions of governance clashed in the room: business lines steering their own transformation, or a central unit orchestrating it. We didn't settle it that day. And that's fine: a disagreement put on the table, named and documented, is worth infinitely more than a fake consensus buried on page 47 of a report.
An AI strategy that holds isn't a document. It's an executive committee that leaves knowing what it has decided, what it still has to decide, and who carries what. The rest — the tools, the models, the POCs — follows.
Geoffrey is Dotika's CEO. He runs days where executives vote more than they listen — and nobody complains.