What is an agent register?
An agent register is a single overview of all AI agents in an organisation: what they do, who owns them, where they run, at which autonomy level and with which risk. An agent that is not in the register does not run.
Swipe to see the full table →
| Agent | Department | Runs on | Owner | Autonomy | Risk | Status |
|---|---|---|---|---|---|---|
| Sort customer questions | Customer service | Copilot Studio | Service team lead | Level 3 | Limited | Pilot |
| Diagnose outages | Operations | Own cloud | Head of operations | Level 2 | Limited | Design |
| Match invoices | Finance | Copilot Studio | Head of accounts payable | Level 2 | Minimal | Shadow |
| Prepare quotes | Sales | Salesforce | Sales manager | Level 2 | Minimal | Pilot |
| Check permit files | Projects | Own cloud | Project lead | Level 1 | Minimal | Design |
- Know what runs where
- Clear ownership
- Ready for audit
- Avoid duplicate work
Explanation
The register is the first thing an auditor, a regulator or a new CISO opens. It is also the natural place to attach obligations from the EU AI Act, NIS2 and DORA.
Every line in the register links to that agent's agent passport. The register gives the overview, the passport the details.
You can start small, with a shared list in Teams or SharePoint. In Agent 365, Microsoft offers an agent register for agents in its own environment. Because most organisations run agents on several platforms, you still need a register that looks across platforms.
The Dutch government already has an algorithm register. An agent register builds on the same idea, with extra attention to what agents are allowed to do on their own.
