Nothing goes to a customer. A simulation never places a real call, sends a text, or books a visit.
Two ways to test
Replay conversations
Run recent real calls through the agent again. Choose how far back to look and how many times to replay each one. Best for checking that a change didn’t break what already works.
Create scenarios
Write the situations your team cares about. Best for something new, like a recall push or a tricky trade-in question.
Run one
1
Pick the agent
Choose which agent to test.
2
Pick the conversations
Replay recent calls, or create scenarios. To start from a common worry, use one of these: First impressions, Getting to the right person, Difficult callers, Inventory uncertainty, Booking a visit, or Callback follow-through.
3
Make the customer harder
Add caller challenges such as Talk over the agent, Correct an earlier detail, Noisy customer speech, Voicemail-like speech, A teammate takes over, or Ask to stop contact.
4
Review and run
Check the setup and start the run. Results fill in as each conversation finishes.
Read the results
Each conversation is checked against a few plain questions:- Did the customer get a useful next step?
- Were answers and promises supported?
- Did the opening match this agent’s setup?
- Was each action, such as a booking or a transfer, handled correctly?
- On calls: did it sound natural and responsive?
Score 45 → 82 · 8 better · 3 same · 1 worse). Open any conversation to read it and see why it was marked that way.
What to test before going live
Fix what you find
Then run the same simulation again and compare. Keep testing after launch: whenever you change instructions or knowledge, replay last week’s calls before customers hear the change.