Skip to interview prep
Apexed

83. An AI expense classifier is 95% accurate. What else would you need to know before using it?

MediumAI & data governance

Technical
Show model answer

Ask how accuracy was measured, which transaction types were tested and whether the sample reflects production. Inspect errors by value, tax sensitivity and account type; a small number of high-impact mistakes can matter more than many easy correct classifications.

Define confidence thresholds, human review and ongoing checks with the accounting owner. Compare against the existing process and include exception-handling effort rather than assuming one aggregate percentage establishes accounting reliability.

What interviewers listen for

  • Inspect test representativeness
  • Weight errors by consequence
  • Design review and monitoring
An AI expense classifier is 95% accurate. What else would you need to know before using it — Finance & accounting interview question — Apexed