How do you brief a leadership team on AI?

01The short answer

Briefing a leadership team on AI does not mean explaining how a language model works. It means showing the team what its own people already do with these tools, helping it weigh what that commits the company to, then having it settle a few questions it had to settle anyway.

The method in short. Start from the reality of the company. Show rather than explain. Set out three issues, risk, compliance and decision. Leave the room with written decisions and a next step under way. The rest is decoration.

Three terms circulate on this subject, and it helps to tell them apart. A leadership briefing aims at the decision. It is for those who decide. AI acculturation of executives is the result of several sessions and repeated practice, a shared language and shared reflexes. AI literacy is the notion European law uses. Article 4 of the AI Act (Regulation (EU) 2024/1689, in force since 1 August 2024) sets an AI literacy obligation, meaning training the people who use AI. A leadership session is a first step. It does not replace training the teams.

Definitions of these terms are gathered in WYP's AI glossary.

02Why this is a leadership team issue

Because use came before decision. Executives often discover AI through the press or a vendor. Their teams already use it. The available data points the same way, with different scopes that should not be added together.

• Adoption. In the United States, 45.2% of employed adults aged 18 to 64 use generative AI for work in Q2 2026, up from 37.4% in August 2025 (FRED, Federal Reserve Bank of St. Louis, value published on 4 August 2026). In France, 40% of micro-enterprises and SMEs say they use at least one artificial intelligence solution (France Num Barometer 2026).
• Unframed use. Two thirds of organisations observe AI use that is not framed by any rules, known as shadow AI (ESSCA and Forvis Mazars barometer on AI and sustainability, 826 respondents, published on 1 October 2026).
• Internal rules. Only 44% of organisations have formalised ethical rules on AI. And 28% of respondents think their company knows the AI Act well (same barometer).
• The cybersecurity view. Employees' use of unapproved AI tools has become the first risky behaviour identified by French CISOs (annual barometer by CESIN and OpinionWay, 397 cybersecurity managers and CISOs, survey from December 2025 to January 2026).

These figures do not say your company has a problem. They say it would be unwise to assume otherwise without looking. Scopes differ, some figures come from the United States, others from respondents working in IT. Keep the common direction, not a percentage to wave around.

03The three issues to put on the table

A leadership session does not have to cover all of AI. It has to install three questions, in this order or another depending on the company.

Risk. Which data leaves the company, towards which tools, through which accounts? Personal use of a consumer tool does not carry the same guarantees as a professional contract. The terms on data use vary with the plan and the account used. The leadership team does not need to know them all. It needs to know that the question exists and who handles it.

Compliance. The AI Act classifies AI systems by level of risk: unacceptable, high risk, limited risk, minimal risk. It is read use by use, not tool by tool. Article 4 sets AI literacy. Article 14 provides for human oversight of high-risk systems. Personal data protection comes on top of this framework. The leadership team does not need to become a legal department. It needs to know which uses deserve a serious look.

Decision. Who authorises a use? Who reviews an output before it goes to a client? Who signs? This is the most underestimated question. AI sharpens the decision, it does not make it. Human oversight is only worth something if it has a name, a person who reviews and puts their signature behind it. The glossary refers to this idea as human in the loop.

The order depends on the house. A services SME will more readily start with data. A group in a regulated sector, with compliance. Decision conditions the other two.

04Preparing the session: start from reality

The session is won before it takes place. Generic slides produce a polite nod. Your own usage produces questions. Prepare the following.

• A map of usage. Who uses which tool, for which task, with which account, personal or professional. A short questionnaire to managers and a few interviews are enough. Also look at the subscriptions the company already pays for.
• Real use cases. A few concrete situations, anonymised, ideally one per department. A well-described AI use case beats a general principle.
• The data involved. What has been entrusted to a tool: e-mails, meeting minutes, customer data, financial information, contract extracts. No names or details, only the nature of the data.
• Decisions already made without saying so. A tolerated tool, an oral rule, a subscription signed by one department. They already exist. Better to make them visible.
• Who is in the room. The leadership team, plus the functions that carry the risk: IT security, legal or data protection, human resources. Also invite a business manager who really uses AI. They make the subject concrete in a few minutes.

If time is short for the usage map, start with the AI maturity diagnostic. It is free and gives a quick picture of the organisation before you go into detail.

05The session agenda, step by step

This is the order that works. The length depends on your team. Keep an ordinary leadership meeting format rather than a seminar. If time is short, keep the live demonstration and the final decisions.

Plan for a facilitator who does not sell tools, an executive who opens and closes the session to give it legitimacy, and someone who records the decisions.

06Common mistakes

• The theoretical lecture. A history of AI, a tour of tools, projections. The team learns, applauds, and decides nothing.
• The tool demo with no decision. The surprise fades within days. If the session produces neither a scope nor a person in charge, it has changed nothing.
• Briefing without looking at usage. The talk stays abstract and the team keeps the impression that the subject concerns others.
• Confusing a leadership session with compliance with Article 4. The AI literacy obligation targets the people who use AI. Briefing the leadership team does not train those who use it.
• Letting a vendor run it. A vendor briefs on its own product. That is not the same as briefing on your situation.
• Writing nothing down. A spoken decision dilutes. A written decision can be checked.

07What WYP does, and what it does not do

WYP.agency is a strategy consultancy for executives, leadership teams and boards of directors. It is neither a marketing agency, nor a tool vendor, nor a software development company. On this subject, the work follows the method the session shows to the leadership team.

WYP cross-checks several AI models (Claude, ChatGPT, Gemini, Perplexity, Mistral, Grok), compares their answers, spots blind spots and contradictions, then a person reviews and signs. AI sharpens the decision, it does not make it.

• What WYP does. Build the usage map with you, prepare and run the session, and record the decisions and open points in writing.
• What WYP does not do. Resell a tool, build one, or promise you that one session is enough to bring the organisation into compliance.

The entry offer is called Le Cadrage IA. It is a fixed-scope package that includes an audit of usage, a written AI policy, a validated tool and a presentation of findings to the leadership team. The team arrives with its own data, which is the best condition for deciding.

08Where to start

• Take the AI maturity diagnostic, free and short, to place the organisation.
• List the uses you already know about and the questions your teams ask you. They are the first cases for the session.
• Schedule the session within an existing leadership meeting, with the functions that carry the risk.
• If the team then wants a written policy and a validated tool, look at Le Cadrage IA.
• For the terms the team will meet (AI literacy, AI Act, shadow AI, AI use case, human in the loop), see the AI glossary.

A successful session shows in this: the team leaves with fewer certainties about AI and more decisions made.