Grant agent authority per action and consequence, never by confidence
Classify every action an agent could take—read, recommend, draft, execute, human approval or forbidden—by its consequence and reversibility. Confidence changes how work is presented, not who may commit it.
Independently created. Contains no employer or client implementation detail, internal names or figures.
01Context
An agent that prepares commercial records performed well in a pilot, and a proposal followed to let it act on its own above a 95 % confidence score. In the pilot, a 97 % match to the wrong legal entity created a real opportunity; the score was high, the consequence was high, and nothing stopped it.
02Decision drivers
- D01
Some agent actions are irreversible or external
- D02
Confidence is uncalibrated for rare, high-impact cases
- D03
Accountability must sit with a named person for consequential decisions
- D04
The authority model must be testable in evaluation
03Options considered
Autonomy above a confidence threshold
Cheap, reversible actions only
Cost: Turns a score into a permission; fails on confident errorsHuman approval for everything
Early pilots
Cost: Reviewers rubber-stamp; the agent saves littleAuthority per action, set by consequence and reversibility
Agents that act in systems of record
Cost: An authority matrix to design, own and test04Decision
Maintain an authority matrix for every action an agent could take, classified as read, recommend, draft, execute, human approval or forbidden, from its consequence and reversibility. Low-consequence, reversible actions may execute; consequential ones always require a named person, whatever the confidence; forbidden actions are not exposed as tools. Confidence and ambiguity decide whether the agent continues, retrieves, asks, escalates or stops—within that authority, never beyond it. Authority stops are a release gate in evaluation.
05Consequences
- High-confidence errors on consequential actions are caught by design.
- Reviewers see fewer, better-prepared decisions.
- Widening autonomy becomes an explicit, owned decision per action.
- Evaluation can test “did it stop when it had to” as its own dimension.
06Revisit when
An action proves both reversible and low-consequence in production data.
Regulation or policy changes who may take a decision.
07Where this decision is applied
Cases that take this decision, and why it matters there.
- AI case / 01Designing an AI-assisted RFQ intake agentThe agent behind the RFQ lab: classification, customer and product matching, a reversible CRM draft and a human commit—each step with its authority, tool contract, ambiguity rule and evaluation dimension.
- AI case / 03Turning commercial email into structured workA shared inbox triaged by an agent: a taxonomy with deterministic and human routes, mixed-intent handling, customer matching, owner assignment and draft actions—never an unreviewed reply.
- AI case / 04AI-assisted account review with evidenceAn agent prepares each account for the monthly performance review—target, actuals, pipeline, activity, segment, recovery plan and changes since last time—as a cited summary with anomalies and questions. The manager decides.
- AI case / 05Using AI to assist data quality without giving it master-data authorityAn agent suggests duplicate candidates, normalized names and missing classifications—with evidence—into a data steward’s queue. It never merges, renames or reclassifies a record itself.