# Stay in Control of Your AI Agents

## Nothing moves without you.

Trust takes time. Your agents earn it by pausing before they act and reporting back when done. Approve, reject with notes, or let timeouts keep things moving.

![Approval inbox showing a pending "Onboard new customer" task with step-by-step review and approve/reject actions](diagrams/control.svg)

### Example: Approval in action

An agent completes the "Onboard new customer" task and pauses for review:

- **Research the customer** *(Sales Dev Rep)* — Finished
- **Draft welcome email** *(Email Marketer)* — Finished
- **Review and send** *(Account Manager)* — Pending approval

The agent presents its work:

> **Onboarding package ready for Acme Corp:**
> - Personalized welcome email drafted with 3 key talking points
> - 30/60/90 day success plan attached based on their growth goals
> - Custom ROI deck generated for Enterprise tier
>
> All materials reference their recent Series B and expansion into EMEA. Ready to send to jamie@acmecorp.com.

You review, then approve or reject with notes.

### Control features

- **Approve before it happens** — Agents pause before calling external tools. Review the action, then approve or block.
- **Review completed work** — Step outputs land in your inbox. Read the results, approve to continue, or reject with notes.
- **Feedback that sticks** — Reject with a reason and the agent retries with your guidance. Fewer corrections over time.
- **Auto-approve timeouts** — Set timeouts so work doesn't stall. If you don't respond, agents continue on their own.
- **Tool-level control** — Disable specific actions or require approval for others. You decide what your agents can do.

---

*© 2026 Agentican Inc. All rights reserved.*