The question before approving an agent isn't "can we build it?". We almost always can. The question is who will make a different decision because of it, and what happens to them if they don't. If that has no answer with a name attached, there is no use case — there's a deliverable.

This session lays out a portfolio model for Microsoft 365 agents: how a proposal is scored before anything is built, the thresholds that let it move to pilot, the evidence to demand before production, and — the part nobody ever designs — how an agent that no longer earns its place gets retired, instead of drifting into a permanent "phase two". We'll look at the four indicators that genuinely predict adoption and the three that only look like they do.

This is not a technical governance session: it isn't about inventorying and securing the agents you already have, it's about deciding which ones should exist at all. You leave with the decision template and the scoring model.