The first-party cast
Decide what to demo from the shelf versus what you built — framed by the success criteria, not feature tourism — and be honest about consumption.
- Choose first-party agents by the success criteria, not by novelty
- Distinguish what is GA from what is preview across the Sales and Service rosters
- Be honest that agent runs consume credits, because Security will ask
Not everything is worth building. The platform ships a cast of first-party agents, and the SE’s judgment is deciding which of them to demo and which to skip. The direction is that role agents move into M365 Copilot⊙ . On the sales side, the Sales Qualification and Sales Research agents are GA, and a Sales Close agent is in preview⊙ . On service, the Case Management, Customer Intent, Quality Evaluation, and Customer Knowledge Management agents went GA in October 2025⊙ , the Customer Service Agent itself is GA as of June 2026⊙ , and Contact Center adds GA voice agents with more in preview.⊙
You choose against the criteria. S2 wants a mandate lifecycle with stage-appropriate AI; S5 wants safe tier-1 deflection. Pick the shelf agents that prove those, and leave the rest out. Feature tourism — showing a capability because it is new — is how a thirty-five-minute demo becomes fifty and lands nothing.
The librarian builds the cheat sheet you actually demo from: for the demo date, which agent is GA, which is preview, and where each is documented. Release waves move features between those columns on a schedule, so this is a standing task, not a one-off — the librarian re-runs it as the date approaches and flags anything that slipped.
The consumption paragraph
A Fortune-50 Security team will ask what these agents cost to run, and “it’s included” is the wrong answer. Copilot Studio meters Copilot Credits, and autonomous agent runs always consume them⊙ . Say so plainly, name the metered actions, and point at the pricing page for the current rates rather than quoting a number that will drift. Consumption honesty is not a weakness in the pitch; it is what makes the governance conversation in the next lesson credible.
In the field
Ask the librarian to build a GA/preview cheat sheet for three first-party agents you are considering. Then map each to a success criterion. If an agent maps to none, cut it from the demo — however much you want to show it.