A diagnostic for deciding which tasks AI agents can perform and with what authority. Each task's authority ceiling is the lower of what the risk allows and what the evidence supports. Every missing piece of evidence becomes an owned action.
What this diagnostic covers
The asset contains:<ul><li>An authority scale for agent work.</li><li>Scoring anchors for failure severity, reversibility, detectability and evidence quality.</li><li>Written formulas for completion, escalation, rework, review time and released capacity.</li><li>A diagnostic questionnaire covering task risk, accountability and approval checkpoints, technical controls, users and baseline.</li><li>An evidence inventory, an evidence clinic challenge log and an assumption register.</li><li>Approval gates, a risk register and KPI definitions with owner roles.</li><li>A synthetic claims-operation example with a failure case, an HR variant and a practice exercise.</li></ul>
- Purpose and decision
- Context and applicability
- Inputs and evidence
- Definitions and assumptions
- Method and analytical logic
- Current-state assessment
- Options and business case
- Recommended design
- Execution and activation
- Governance and risk
- Measurement and benefits
- Worked example and practice
- Working tools and references