A new book from Harvard Business Review Press makes a simple but pointed argument: most companies are using AI in meetings the way they would use a tape recorder, and that is a waste.
The authors, both from Capgemini's consulting arm, point to research showing a gap. Capgemini's own 2024 survey found that only 15 percent of managers used generative AI daily. That number has since grown to roughly half of managers, according to the book. But only a small fraction, around 7 percent, use it for team activities such as meetings, planning sessions, and workshops. Most people still treat it as a personal writing tool, not something that sits in the room with the team.
That gap matters more than it sounds. Meetings, workshops, and planning sessions can eat up a third or more of a typical work week. If AI is only helping with solo tasks like drafting emails, it is missing the part of the job where a lot of time actually goes.
What makes this worth paying attention to now is that the idea is not stuck in a book. It is already built into tools many companies already pay for. Microsoft added a feature to Teams called Facilitator that builds meeting agendas, tracks time on each topic, and writes up decisions as the conversation happens. Zoom and Google Meet both rolled out similar meeting assistants of their own this year. So the technology to do exactly what this book describes is sitting inside software licenses that most office workers already have.
The book pushes the idea further than just note-taking. It suggests AI can be assigned a role in a meeting, such as playing devil's advocate, standing in as a difficult client, or acting as a rival company during a planning exercise. The goal is to catch bad decisions before they happen, the same kind of failure that historians point to in disasters like the Bay of Pigs invasion, where a group agreed too quickly without anyone raising the obvious objections.
There are real risks worth naming plainly. AI can get facts wrong, and letting it sit in on sensitive planning conversations means company information is now flowing through another system. Anyone doing this needs basic rules: check what the AI concludes before acting on it, and be careful what gets fed into it.
For a business leader deciding where to spend limited time and budget on AI, this is a useful signal. Companies that are not ready for the harder work of automating entire processes with AI can start somewhere much simpler and cheaper: turning on the meeting features already included in software they are paying for anyway, and treating team meetings as the first place AI earns its keep.