Strategic keynote

Deciding with AI without giving up your judgement

AI produces answers in seconds. It does not carry the responsibility for you.

The problem is no longer getting an analysis. It is knowing what it leaves out, what makes it credible, and at what point a human has to slow down enough to take back control.

In McKinsey's 2025 survey, 39% of respondents attribute an impact on their organisation's EBIT to AI, most of them placing it below 5% of that EBIT. As AI enters the results, the quality of decisions and controls becomes a leadership question.

McKinsey, The State of AI in 2025

More analysis does not mean better decisions

Papers are summarised faster, scenarios presented better and recommendations made more convincing. The time available to decide does not grow. A fluent answer can then be mistaken for a solid one.

As more steps of the reasoning are delegated, another risk appears: no longer being able to explain how a conclusion was built. Yet a decision belongs to the person who signs it, even when it leans on a machine.

Deciding when everything speeds up does not try to slow the organisation down. It teaches it where to accelerate, where to verify, and where to keep an irreducible amount of human time.

The question it addresses

When AI multiplies the analysis and accelerates the choices, how do you preserve judgement, courage and accountability?

Who it is for

Leadership teams, management offsites, director meetings, leadership clubs and professional congresses.

Decision
Judgement
Courage
Prioritisation

Choose it if your teams have more information than ever but struggle to spell out their criteria, their trade-offs or their level of confidence.

What the keynote sets in motion

Recognising the decisions that require slowing down

Not every important decision is alike. The keynote separates those where AI usefully sheds light, those where it can hide uncertainty, and those where the consequences demand stronger checks.

Putting three questions back before the call

What is this recommendation based on? What is missing from it? What would we know if we had built the reasoning ourselves? Those simple questions change the quality of a conversation.

Making accountability explicit

Saying who decides, who checks and who answers for it does not slow action down. It lets you use AI with more composure, and with vigilance proportionate to the risk.

A decision replayed, not a lecture

The heart of the talk can take the form of a replay. A decision the room knows is examined once from the available recommendation, then a second time with the assumptions, the missing data and the discarded alternatives made visible.

The point is not to prove that humans are always right. It is to show what the decision process must preserve once the machine accelerates the production of answers.

Ground already covered on stage

At the Cartes Bancaires CB connected commerce day, Salime spoke on “Le commerce de demain : perspectives et prospectives”, on the future of commerce. That reference attests to public work on anticipation and choices in a shifting technological environment.

Formats

45 to 60 minutes. Up to 90 minutes with a decision exercise. French or English. For a small committee, internal cases can be prepared subject to framing beforehand.

The questions that come up

Is this a talk against AI?

No. It is aimed precisely at organisations that want to use AI more, without confusing speed, confidence and decision quality.

Does the audience have to be senior leaders?

Not necessarily. Everyone decides at their own level. The examples and the degree of responsibility adapt to the room.

How does it differ from Leading the augmented company?

Leading the augmented company is about governance and what can be delegated. Here the subject is the quality of the reasoning once AI has already produced an analysis or a recommendation.

What does it cost?

It depends on the format, the location, the language and the preparation work. If an internal case is to be used, that level of tailoring is built into the proposal.