L0 · Observe
Record results without taking action.
OPRYNA · AI WORKFORCE
AI Agent Governance
AI Agent Governance keeps authority with the company. Each process receives an agreed L0–L4 level; pilots start at L0–L1, and higher authority requires company approval.
Five levels of authority, agreed per process
L1 · Recommend
Provide suggestions; the team decides and acts.
L2 · Prepare
Prepare transactions or messages for approval.
L3 · Act within limits
Execute only within company-approved limits.
L4 · Operate autonomously
Complete a fixed process scope while the team audits.
WORKING TOGETHER
A clear starting point
Authority applies per process, not per agent. The company may lower it at any time. Written business rules, restricted access, audit evidence, and human exception review define the operating boundary.
Review authority and approvals →
Explore related work
AI Agent Governance: frequently asked questions
What does AI Agent Governance cover?
AI Agent Governance keeps authority with the company. Each process receives an agreed L0–L4 level; pilots start at L0–L1, and higher authority requires company approval.
Who approves the agent’s actions?
Your company agrees authority for each process. Pilots start at L0–L1; higher authority requires approval and can be lowered at any time.
What should we prepare first?
Prepare one process owner, an IT contact, and 20–30 document samples. An NDA precedes company data transfer; samples may be anonymized.
For an external AI governance reference, see NIST AI Risk Management Framework.
Choose one process to discuss
Map the work, confirm evidence, and agree what a useful pilot should measure.