Approvals
Use approvals to review sensitive actions before Operator runs them.
Approvals let Polaris Operator prepare an action, but wait for human confirmation before running it.
The flow is simple: AI proposes, a person reviews and Polaris executes only after approval. This lets automation and human supervision work together.
When to use approvals
Use approvals when an action may affect a customer, an important process or a sensitive decision.
Examples:
- Sending sensitive messages.
- Approving discounts or special terms.
- Changing the status of an appointment, lead or request.
- Running an action in an external tool.
- Launching an action for many users or conversations.
How they work
Operator prepares the action with the available context. Then a person can review it, approve it or reject it.
Polaris records the result so the team can see what was requested, who reviewed it and what happened next.
How to configure them
- Identify sensitive actions.
- Define who should review them.
- Decide what information must be shown before approval.
- Test a real request.
- Enable approvals before automating that action.
Approval examples
| Action | Recommended approval |
|---|---|
| Confirm appointment | Yes |
| Create lead | Optional |
| Send sensitive message | Yes |
| Create internal task | Not always |
| Apply discount | Yes |
Work continuity
Even if a workspace reaches certain usage limits, Polaris keeps approvals accessible so the team can review or close open work.
This prevents important automation from getting stuck without visibility and helps your team stay in control.
First recommendation
Start with approvals for any sensitive action. Once you have enough history and confidence in the flow, you can decide which actions are safe to automate with less review.
What to avoid
- Approving without enough context.
- Automating sensitive changes on day one.
- Not defining who reviews approvals.
- Leaving requests open without follow-up.
Related modules
Approvals protect Actions, Booking, WhatsApp and Operator flows.
They are especially useful when AI can prepare work, but a person should confirm before affecting a customer or important process.