Skip to course content
Free agentic AI fundamentals course

Introduction to Agentic AI and Workflow Automation

Unit 06.02: Making a rejection worth the reviewer's time

A rejection with no information costs a full run and produces the same output again.

Three fields make it actionable

What was wrong, what the correct value is, and what to do next.

The table below compares three rejections.

REJECTION                                                        USABLE?
(blank)                                                          NO
"not right"                                                      NO
"Wrong amount. The correct figure is 120.00. Re-check the
 payment log for partial refunds."                               YES

  Three fields make a rejection actionable: what was wrong, what the correct
  value is, and what to do next. Give the reviewer a form with those three
  boxes rather than one free-text box.

"Not right" is a rejection the system cannot act on. The retry has exactly the information the first attempt had, so it produces exactly what the first attempt produced - at full cost.

Structuring the reviewer's form around the three fields is what makes this reliable. A free-text box produces "not right" because that is what is quick to type; three labelled fields produce three answers because empty fields are visible.

The mistake this prevents

The mistake is measuring the rejection rate and not what follows one. A workflow where every rejection leads to an identical retry has a rejection rate that looks like quality control and functions as a way of paying twice for the same output.

Takeaway

Structure the rejection into what was wrong, the correct value, and the next action. A free-text box produces rejections nothing can act on.