Where in an agent workflow should a human approval gate be placed for an irreversible action?

Attempt it

Quick reference

TL;DR

Gate irreversible actions before they execute. A post-action review cannot undo a sent email or a deleted record, so the human must approve while the action can still be stopped.

Memory aid

Sign in for the mnemonic that makes this stick. See Pro pricing.

Easy to grasp

Imagine your assistant is about to mail a letter for you. If they ask 'should I drop this in the mailbox?' before they do it, you can say no and nothing bad happens. If they mail it first and then tell you, the letter is already gone and you cannot get it back. Some actions are like dropping a letter in the mailbox or deleting a file: once they happen, you cannot take them back. For those actions, the assistant has to stop and ask you first, while there is still a chance to say no. For small, easy to undo things, like writing a draft you can still edit, the assistant does not need to ask. The rule is simple: stop and ask before anything that cannot be reversed.

Key concepts

Human-in-the-Loop Action Reversibility Blast Radius

Detailed answer & concept explanation

~7 min read

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example. Click to expand.

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example.

Sign in for 5 free unlocks/day or go Pro for unlimited. See Pro pricing.

Quick scan · 2–4 min

Everything important, quickly.

State that irreversible actions are gated pre-execution, justify it by timing, then generalise to the reversibility times impact axis. Cover where the classification lives, how the loop interrupt works, and the refinements of preferring reversibility and logging every approval for audit.

Real-world examples

Real products, models, and research that use this idea.

LangGraph exposes human in the loop as an interrupt that pauses before a node, persists state to a checkpointer, and resumes only when the human approves the proposed tool call.
Claude Code prompts for explicit approval before running shell commands or editing files, but auto-allows read-only actions, which is reversibility-keyed gating in practice.
Anthropic's computer use guidance recommends a confirmation step before consequential clicks like sending or purchasing, while leaving navigation and reading ungated.
Stripe-style payment agents gate the charge call behind explicit sign-off while letting balance reads and quote generation run freely, since a charge is irreversible and a read is not.

Interviewer follow-ups

What an interviewer would ask next. Try answering before peeking at the approach.

QHow would you classify which tool calls require a pre-execution gate versus auto-approval?

Sign in or go Pro to see the approach for this follow-up.

QHow do you stop approval fatigue from making the human a rubber stamp?

Sign in or go Pro to see the approach for this follow-up.

QWhat changes when a single human cannot keep up with a fleet of agents needing approvals?

Sign in or go Pro to see the approach for this follow-up.

Don't say this

Red flags and common mistakes that signal junior thinking. Click to expand.

Most common mistake

Placing the approval gate after execution so the human can review what happened. For an irreversible action that is too late, because the damage is already done and the review cannot undo it.

Sign in or go Pro to see all red flags and common mistakes. See Pro pricing.

60-second cheat sheet

The night-before-the-interview bullets. Scan these on the way to the call.

Sign in for 5 unlocks/day or go Pro for unlimited revision sheets. See Pro pricing.

References

Primary sources. Skim if you want the original framing.