The Cost of Asking: Team Practices in the Age of Agentic AI
- Philippe Beaudoin
- November 17, 2026
- Vector Institute, Toronto
Asking a teammate used to cost something: a small interruption, a moment of admitting you didn’t know. Agents made that cost disappear, and with it, quietly, some of the conversations that held teams together. Nobody decided this. It simply happened, the way most changes in a team’s culture happen.
This talk, which opened a day of reflection for the Vector Institute’s AI Engineering team, looks at that shift through practices: the observable things we do together, again and again. Whether we ask our teammate or our agent first. Whether we say good morning. Whether we apologize when we’re wrong. Our working hypothesis is that what we feel about our work, and about each other, grows out of these practices, and in turn shapes the next ones. Practices can be named, examined and changed without first agreeing on a theory of AI, or on how anyone should feel about it.
Some practices we keep on purpose. Some we are still reaching for. And some we adopted without ever wanting to. Seeing which is which is something a team can only do together.
I also introduce the work we’re beginning at Intersubjective Labs with Potted Plant: an open, collaborative commons where teams that include both people and AI systems can share and study their practices, the way developers share packages. It holds no opinion about which practices are good. The aim is to help teams discover, in their own context, what keeps them cohesive, curious and effective as their tools and roles keep changing.