Glossary term
Glossary term
Product and Operations
Governance artifact defining allowed autonomy levels by task risk.
A global insurance company created an Agent Autonomy Matrix mapping 40 process steps across four autonomy levels: 'AI suggests only', 'AI drafts, human reviews', 'AI acts, human monitors', 'AI acts autonomously' - approved by the risk committee.
A regulatory body's AI governance policy mandates that any agent action affecting citizen rights be classified at Autonomy Level 1 (human decides) - the matrix is reviewed annually and updated when new agent capabilities are deployed.