Skip to content
RED SAGSkills
hardBusiness Analyst interview question

How would you write requirements for a feature that uses AI, such as automatically classifying customer emails?

What the interviewer wants to hear

  • Define the business outcome and success metrics
  • Specify acceptable accuracy, such as precision and recall per class
  • Define fallback and human review for low-confidence cases
  • Cover data needs, privacy and bias
  • Plan monitoring and retraining after launch
  • Set expectations that results are probabilistic

Model answer

I would start with the business outcome, such as cutting the time to route emails from hours to minutes, and how we will measure it. AI outputs are probabilistic, so instead of 'the system classifies emails correctly', I would define acceptable performance, for example 90% precision for the complaints category, measured on a labelled test set. I would define what happens when the model is unsure: emails below a confidence threshold go to a human queue. I would cover data requirements, including where training data comes from, privacy rules for personal data and checks for bias, such as language or region. There would be requirements for explaining or logging decisions where needed. Finally, I would specify monitoring after launch, such as tracking accuracy and drift, and a process for feedback and retraining. I would make sure stakeholders understand some errors are expected and plan for them.

Can you answer it under time pressure?

Take a timed hard round and get scored on every answer.

Start mock interview

More hard Business Analyst questions