Each task is one governed session: who ran it, on which model, under which policy, with what outcome. It writes itself, so nobody has to remember to document anything.
The Governed Session
Not a log you export at the end of the quarter, but a record written as the work happens. See the AI audit trail.
In Plain Language
A team member asked for a rent-increase notice. Because it involved confidential tenant data, policy required an approved frontier model. It ran in 4.2 seconds. The result was accepted. No overrides.
Policies
Start from an industry pack, describe the rule in plain English, or build a WHEN / THEN rule from scratch. Nothing is live until you simulate and publish. See industry packs.
Admin Oversight
One place to see sessions, cost and policy activity, and to change a rule that applies to the next task. See the admin dashboard.
1,284
governed sessions this month
214 / wk
hits on the busiest rule
0
unlogged overrides
$38 / $100
org monthly budget
Illustrative sample data.
Evidence a regulator or auditor can open, with retention of 90 days on Pro and a year or more on Enterprise.
Every action has a tier. Tier-3 actions need a human before they run, and every override is logged.
Caps per team and per user with alerts at 50, 80 and 100%. Agent tasks get a hard ceiling. See RenOS Credits.
One task from request to accepted result: who ran it, on which model, under which policy, with what outcome, written as the work happens.
Each record is chained to the one before it. Any change breaks the chain and shows up on verification.
Every rule is linted and simulated against the last 30 days of sessions before it's published. Lint and simulation are deterministic; no model grades its own policy.
90 days on Pro, and a year or more with export on Enterprise.
Book a pilot and open a governed session from your own workflow.