Live reactive voice test
What should we improve first?
What should the agent do after it receives this answer?