Review fatigue is the gradual drop in care that sets in when a person reviews a long run of work that is mostly correct. Each approval becomes quicker and lighter, until the reviewer is confirming rather than checking. It is a particular risk in AI agent work, because agents produce high volumes of drafts that are usually right.
Why does review fatigue matter?
The whole case for letting agents draft while people approve rests on the approval being real. Review fatigue hollows that out without anyone noticing. Every action still carries a named approver; the record looks perfect. But the approver stopped looking somewhere around the eightieth case of the afternoon.
Employers hiring reviewers know this risk well. A candidate who can explain review fatigue, describe its signs and say what they do about it is showing exactly the self-awareness the job needs.
How does review fatigue show up in practice?
It shows in behaviour before it shows in errors:
- Decision times shrink steadily through a session.
- You stop opening the source evidence and rely on the draft's summary.
- Rationales become identical: "approved, checks complete".
- You approve a case and cannot recall a single detail about it a minute later.
Habits help. Read the evidence before the draft. Take a short break at fixed intervals. Write a specific reason for every decision, which forces at least one real look. Mix harder cases into routine runs so attention does not settle. And watch your own pace: a sudden speed-up is a warning, not a win. Keeping a decision log for practising AI agent review lets you see your own pattern.
What is an example?
A reviewer we will call Mateo approves agent-drafted replies to delivery complaints at an online retailer. Mornings go well. By mid-afternoon his average time per case has halved. He approves a reply offering a standard apology to a customer whose message, read properly, says the courier left a parcel containing medication on an open doorstep in the rain. That case needed a replacement sent urgently and a note to the courier team. A quality check the next day catches it. Mateo changes his routine: ten-minute breaks every hour and the customer's message read in full before the draft.
How is review fatigue different from alert fatigue?
Alert fatigue comes from a flood of signals, most of which are false alarms; people start dismissing alerts. Review fatigue comes from a flood of drafts, most of which are correct; people start approving them. One leads to missed alarms, the other to approved mistakes. The first is explained in what alert fatigue is.
Review fatigue also makes other biases worse, especially automation bias and anchoring on the draft. The wider family is covered in what bias in review is.
Which terms are related?
- Over-trust: believing the agent is more reliable than it is, which fatigue speeds up.
- Rubber-stamping: approval without real review, the end point of fatigue.
- Reviewing an agent's decision: the task itself, set out in what it means to review an AI agent's decision.
For the skills employers test, read AI skills for job seekers.
Where the Academy fits
The Certified Pryme Operator practicum lasts two weeks, not one sitting, so it measures whether your checking holds up through ordinary working days. The standard is nineteen in twenty correct and no missed red line, signed off by a Pryme assessor. The Operator track at Pryme Intelligence Academy also has a module on running the rhythm of a live workspace day to day.
Questions people ask
How many cases can one person review well in a day?
It depends on the complexity of the cases and the stakes, so there is no single number. Watch your own signals; if accuracy on a sample falls late in the day, the volume or the breaks need to change.
Can I practise resisting review fatigue before I have a job?
Yes. Run long practice sessions on realistic cases, time yourself, and score your accuracy by hour. You will see your own pattern and can test which habits help, as described in how to practise AI agent review before you have a job.