Agentic Systemsintermediate13 min · 9 steps
Human-in-the-Loop Patterns
Put people where they add judgment: breakpoints on consequential actions, policy gates for the rest, approvals bound to what actually runs, and escalation driven by calibrated scores.
requesttool callsignaldataerrorresponse
ProblemStep 1 of 9
Approve-everything turns humans into rubber stamps
The naive design pops an approval dialog for every action. After the fiftieth harmless ls, reviewers stop reading and start clicking.
It's measurable. In Anthropic's blind study of 1,053 paid testers, humans caught a swapped-in dangerous command only 13.6% of the time; the auto-mode classifier blocked 89%.
Watch outPer-action approval doesn't scale — and when humans do review everything, they're a weak filter.
Lesson 8 of 15 in Agentic Systems