PSRM for Agents

Govern the agent before it touches the decision.

PSRM for Agents explains how an organisation decides where AI agents belong in procurement and supply, and what record governs any approved agent use.

Start

The first question

The starting question is whether an agent should be used at all. PSRM maps process, data, decision rights, evidence and controls before agent use is approved.

Outcomes

Four outcomes

Automate under controls

Approved for agent operation, with defined controls and monitoring.

Augment

The agent senses and proves, a human decides.

Keep human

The decision remains with a named human owner.

Fix first

The process, data or decision rights need uplift before agent use.

Pack

The Agent Governance Pack

Where agent use can be governed, the decision is recorded in an Agent Governance Pack. The pack records what an agent may do, how it proves what it did, and when a human must decide.

Permission boundary

What the agent may and may not do.

Evidence schedule

What proof it records and when.

Escalation triggers

When the human must decide.

Drift register

How change is detected and logged.

Licence and provenance record

Sources, models and rights.

Expiry date

When the pack must be reviewed.

Permissions

Permission levels

Observe, Recommend, Execute within bounds.

Rule: A human remains accountable for the outcome.

Tools

Tool boundary

PSRM gives the governance language. Tool choice remains with the organisation. PSRM does not endorse a tool, vendor or catalogue through this page.