← sources

At what point should an AI agent stop and ask for human approval?

source: reddit (reddit:AI_Agents)published 2026-09-29 13:26ingested 2026-09-30 02:37

retained text

I've been thinking about this while building around AI agent security.

Giving an agent access to a tool is relatively easy.

The harder question is: what should that agent actually be allowed to do at runtime?

For example, an agent might be allowed to:

read an invoice

create a draft

update customer information

But what about:

sending money

changing bank/account details

exporting customer data

deleting records

performing a chain of actions that becomes risky only because of the sequence?

I don't think the answer is simply "give the agent fewer permissions."

It feels more like we need to evaluate the action + context + risk + what the agent has already done, and sometimes stop the workflow for human approval.

Curious how others are approaching this.

Where would you draw the line between autonomous execution and mandatory human approval for AI agents?

document fetched 2026-09-30 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1wta2q4/at_what_point_should_an_ai_agent_stop_and_ask_for