Skip to content
ML Academy

RAG and agents

Interview question

  • seniorlevel
  • 12 minsuggested time
  • 2Applied AI, ML engineer

A team proposes an autonomous agent to process flight-disruption refund requests end to end. The process has clear eligibility rules but messy inputs (free-text emails, attachments). Would you build an agent or a deterministic workflow? Explain the design.

Tests: Agents vs deterministic workflows: where LLM flexibility adds value, reliability and testability, cost, and designing bounded LLM steps inside a workflow.

Sign in to practise: answers are saved before model answers are shown.