AI Governance
What it does
A running record of what your AI agents did: the action, its status, which agent took it, the model it used, and when. Open any entry to see the details on record, such as the prompt, the output, and the outcome. Together with the Approvals tab and Agent Autonomy settings, it lets you decide what agents may do and check what they did.
Who can use it
- Only organization Owners and Admins can view the activity log. Anyone else who reaches it sees "Only organization owners and admins can view the AI activity log."
- Only Owners and Admins can change Agent Autonomy and decide approvals.
Where to find it
Click Chief of Staff in the sidebar, then Settings, then the AI Activity tab.
The same log is also available as a standalone AI Governance page, which you can reach from a direct link. That page adds two shortcuts: Approval Center and Autonomy settings.
Read the activity log
- Open the AI Activity tab. The newest actions appear first.
- Each row shows:
- Action — what happened, in readable words, with the system name below it.
- Status — how it was authorized (see below).
- Agent — the agent that acted.
- Model — the AI model it used, where one was recorded.
- When — how long ago it happened.
- The log loads 25 entries at a time. Click Load more to see older ones. The "Showing" line above the button tells you how many of the total you've loaded.
On a phone, Agent moves under the action and Model is hidden. Open an entry to see both.
Filter by status
Use the buttons above the log:
- All — everything.
- Auto-ran — ran without asking.
- Approved — someone approved it first.
- Rejected — someone declined it.
- Pending — still waiting for a decision.
Entries can also show Expired (nobody decided in time), Cancelled (the request was withdrawn), or Logged (recorded without one of the statuses this screen names). These have no filter button of their own, so look for them under All.
Inspect one action
Click any row. The details window can show:
- When — the exact date and time.
- Approval and Agent job — reference IDs that link the entry to its approval or background run.
- Sources — what the action was based on.
- Outcome — the final result.
- Output — what the agent produced.
- Prompt — the instruction or inputs on record.
Only what was recorded appears. If there's nothing more, you'll see "No further details were recorded for this action."
Change what agents can do on their own
- Open Chief of Staff, Settings, then the Agent Autonomy tab.
- Each capability is labeled External or Internal and has a dropdown:
- Ask every time — "You see the action first — a card in chat or voice, or an entry in Approvals for background agents — and nothing happens until you approve."
- Run without asking — "It runs on its own and is logged in AI Activity. You can undo within 10 min where supported."
- Deleting is locked to Always asks: "Some actions can't be automated. Deleting always asks."
- Some capabilities also have Per-action overrides, so you can treat one action differently from the rest. Use Filter actions to find it.
The same rule applies everywhere: chat, voice, briefing Do it, routines, background agents, and connected AI apps.
Decide what's waiting
Open Chief of Staff, then Approvals. Requests you approve or reject there show up in the activity log afterwards. For the full walkthrough, see Chief of Staff.
Tips & limits
- The log records actions, not every question. Writes, approval decisions, and background runs are logged. A chat lookup that only reads data doesn't add an entry.
- Writes made by connected AI apps, such as Claude or ChatGPT, are logged too. See Connect Claude, ChatGPT and other AI tools.
- Secret values stored through the Credentials Vault tools are removed from what gets logged.
- The log is read-only here. This screen has no edit, delete, or export button.
- The log covers one organization at a time.
Frequently asked questions
Who can see this log?
Organization Owners and Admins. Other roles can't open it, even if they can use AI features.
Why does an entry say "Logged"?
It was recorded without one of the statuses the screen names, such as Approved or Auto-ran. Open the entry to see what's on record. An Approval reference in the details means it went through the approval process.
Can an agent act without leaving a record?
The Agent autonomy settings put it this way: "Every action, automatic or approved, is audit-logged." Chat lookups that only read data aren't listed, because they don't change anything.
Why is the Model column empty for some actions?
Not every action calls an AI model directly, and some don't record one. A dash means no model was recorded for that entry.
How do I stop an agent from doing something without asking?
Set that capability to Ask every time under Agent Autonomy. New actions will wait for approval from then on. Past entries stay in the log.
Where do I approve a pending action?
In Chief of Staff, under Approvals. The activity log is for reviewing, so it has no approve button.
Still need help?
Our team answers every message. Already a customer? Open a support ticket from inside Nautis.