When an AI action exceeds your autonomy threshold, Nova does not run it. Instead it waits in the Approval Queue with everything you need to make the call: what it wants to do, why, the blast radius, and the full reasoning behind it.
Go to the Approval Queue. Each card is one proposed action waiting for a human decision.
See exactly what Nova wants to do, on which target, and why. The blast radius tells you how much is affected if it goes wrong.
Open the decision bundle to read the full reasoning transcript and the data Nova used. This is the same record auditors see.
Approve to run it now, edit the parameters first, or reject with a reason. Rejections feed back into the agent trust score.
Set how long an approval can sit before it escalates to the next person, so nothing stalls during an incident.
Approvals are single-use (protected by a replay guard) and every decision is recorded for SOC2 and HIPAA audit. Pair this with the AI Activity Timeline for the complete picture.