How do I make automations and AI agents ask for human approval?

Put an approval step in front of every action that is hard to undo: payments, refunds, messages to customers, deleting or overwriting records, anything above an amount you set. The workflow pauses, a person gets the details by email, Slack or chat, and the run continues only on approve. Zapier, n8n and Power Automate each have a built-in way to do this.

Updated 8 November 2026.

Ask us about your case See use cases

This page answers

How can I make sure my AI agents know when to ask for human approval?
How do I create automated workflows with conditions and approvals?
Can I require approval before AI sends payments or modifies sensitive records?
Can I prevent AI agents from performing unauthorized actions?
How do I automate internal approval processes?

What should need approval

Action	Approve?	Why
Read data, draft a reply, prepare an entry	No	Nothing leaves or changes
Update an internal field, add a tag	No, but log it	Easy to see and undo
Send a message to a customer or supplier	Yes at first; later only for new cases	Cannot be unsent
Create or change an order, invoice or price	Above an amount you set	Money and records
Payments and refunds	Always	Hard or impossible to reverse
Delete or overwrite records	Always	Data loss

How the main tools do it

Tool	Built-in step	What the reviewer can do
Zapier	Human in the Loop app, Request Approval action	Approve, decline, or edit the data before the Zap continues; on decline the Zap either continues or stops, as you choose. Notifications by email, Slack or another Zap. Zapier lists it as a premium app that needs a paid account.
n8n	Human review on an AI Agent's tools; Wait node	For AI agents: approve or deny a specific tool call before it runs, through Slack, Telegram or the n8n Chat. For any workflow: the Wait node pauses until a webhook call or a form is submitted.
Power Automate	Approvals	Start an approval inside a flow and continue on the response.
Any tool	Link with a token, then a webhook	Send the reviewer a message with Approve and Reject links; the link calls a webhook that resumes or stops the run.

Rules that make approvals work

Show the reviewer everything needed to decide in the message itself: what, how much, for whom, and why the system wants to do it.
Set a deadline and a fallback: if nobody answers in 4 hours, remind; after 24, escalate or stop.
Approve the exact data that will be used. If the data changes after approval, ask again.
Give the agent its own account with only the rights it needs, so a missing approval cannot be bypassed by a direct action.
Log who approved what and when, next to what the automation did.
Start strict, then remove approvals for case types that have been approved without changes for weeks.

Sources

Vendor facts on this page come from these pages, checked on 8 November 2026.

Questions

Can an AI agent decide by itself when to ask?

Do not leave that to the model. Put the rule in the workflow: these actions, above this amount, always wait for a person. The model can add a request for review when it is unsure, but it should not be able to skip a required one.

Does Make have an approval step?

You can build one with the link-and-webhook pattern above: the scenario sends the approval message, and a second scenario started by the webhook carries on.

Will approvals slow everything down?

Only the cases that need them. Routine low-risk actions run straight through; that is the reason to decide per action, not per workflow.

Agents for these jobs

Want an agent built for this?

We build AI agents in code, connected directly to your systems, with no per-task fees. Write a few lines about the work you want to automate and the systems you use; we reply to set up an online intro call.