Series

4 parts

Agent governance in Microsoft 365

Four parts on tools, lifecycle, runtime and evidence, written from the perspective of an operator in regulated environments.

Agents in Microsoft 365 are built with three tools: Agent Builder, Copilot Studio and Foundry. Microsoft distinguishes three models of who carries the risk: the platform, the administrator or the organization. That distinction runs through all four parts of this series.

I write from the seat of an operator in regulated environments (critical infrastructure, NIS2), usually with several tenants and an IT service provider. So each part also asks a question Microsoft does not: how is responsibility split between customer and service provider?

The four parts of the series along risk ownership: part 1 introduces the three models, part 2 shows the lifecycle per model, part 3 how responsibility shifts at runtime, part 4 the evidence

  1. Part 1

    Agent Builder, Copilot Studio or Foundry: who owns the risk?

    Which tool fits which builder, and who carries the risk in each case: the platform, the administrator or the organization?

  2. Part 2

    Agent lifecycle: how much governance does a Copilot agent need?

    How much lifecycle does an agent need, and how does it differ by platform?

  3. Part 3

    Runtime agent governance: whose permissions does a delegated agent use?

    Whose permissions does an agent use at runtime, and where does responsibility shift?

  4. Part 4

    AI Act, GDPR, NIS2 and ISO: which controls for Microsoft 365 agents count more than once

    Who has to provide which evidence, and how much of it comes from Microsoft 365?