Approval fatigue.
Also known as: Rubber-stamp fatigue
When an AI agent asks a person to approve so many things that they stop reading and just click yes. The human is still technically in the loop, but they’ve stopped judging what’s in front of them. Adding more approval steps to an agentic workflow can end up giving you less oversight and judgement, not more, because the human isn’t really in the loop. A bit like cookie banners. When did you last read one of those?
In the wild
“approvals can degrade under time pressure or approval fatigue, creating a false sense of safety”
Seen here: Shiva Varma, Gartner, May 2026
For AI users
Slow down when an agent asks for permission over and over, or asks for something unclear. Read what it’s about to do and what happens if it goes wrong, rather than hitting the button you always hit.
For business leaders
Don’t treat a human approval step as a control just because it exists. Tier approvals by risk so people see the ones that matter, give them enough context to decide properly, and check a sample afterwards. Watch out for approvals that happen suspiciously fast. A log can show someone clicked, but it can’t show they thought about it.