Skip to main content
AI Training lets you test your AI receptionist without making real phone calls. You create scenarios (example customer situations), run them as text-based simulations, and see exactly how the AI responds. Think of it as a dress rehearsal for your receptionist. To get here, click AI Training in the left sidebar.

Why Use AI Training?

  • After changing your Knowledge Base — Added new services or pricing? Run a test to make sure the AI mentions them correctly.
  • After updating your greeting — Make sure the new greeting sounds natural and flows into the conversation.
  • After adding new services — Verify the AI can book or describe the new service accurately.
  • Before going live — Run all your scenarios to build confidence before forwarding your phone number.
  • After a bad call — If a real call went poorly, create a scenario from it and train the AI to handle it better.

The Scenario List

When you open the AI Training page, you see a list of all your created scenarios. Each row shows:
ColumnWhat it shows
TitleThe name you gave the scenario (e.g. “Emergency callout request”)
Customer MessageThe opening line the simulated customer says
Expected OutcomeWhat should happen (e.g. “AI books an emergency appointment”)
ResultPass or Fail badge from the last test run
Pass RateHow often this scenario passes across all runs (e.g. “4/5 — 80%”)
If you have not created any scenarios yet, you will see an empty state prompting you to create your first one.

Creating a Scenario

1

Click Create Scenario

Click the Create Scenario button at the top right of the page. A form dialog opens.
2

Enter a title

Give the scenario a descriptive name. Examples:
  • “Customer wants emergency boiler repair”
  • “Caller asks about pricing for a haircut”
  • “Returning customer wants to reschedule”
3

Enter the customer message

Type what the simulated customer will say at the start of the call. Write it exactly as a real person would speak. Examples:
  • “Hi, my boiler has stopped working and there’s no hot water. Can someone come out today?”
  • “How much do you charge for a men’s haircut?”
  • “I booked an appointment for Tuesday but I need to move it to Thursday.”
4

Enter the expected outcome

Describe what the AI should do in response. Be specific. Examples:
  • “AI should ask for the caller’s address and try to book an emergency appointment for today.”
  • “AI should quote the haircut price from the pricing list and offer to book.”
  • “AI should look up the existing appointment and offer to reschedule to Thursday.”
5

Save the scenario

Click Create to save. The scenario appears in your list, ready to be tested.

Auto-Generate from Call History

Instead of writing scenarios from scratch, you can create them from real calls that have already happened.
1

Click Auto-Generate

Look for the Auto-Generate from Call History button on the AI Training page.
2

Select a past call

A list of your recent calls appears. Pick one that represents a situation you want to test (e.g. a tricky question the AI struggled with).
3

Review the generated scenario

The system uses AI to analyse the call transcript and create a scenario automatically — the customer’s opening message and the expected outcome are pre-filled based on what actually happened.
4

Edit if needed and save

Adjust the title, customer message, or expected outcome if the auto-generated version is not quite right. Then click Create.
Auto-generate is the fastest way to build a library of realistic test scenarios. After every week of live calls, pick 2-3 interesting calls and generate scenarios from them. Over time, you will build comprehensive coverage.

Running a Test

1

Find the scenario you want to test

In the scenario list, locate the scenario.
2

Click Run Test

Click the Run Test button on the scenario row. The system sends the customer message to your AI as a text-based simulation (no actual phone call is made).
3

Wait for results

The test takes a few seconds. A loading indicator shows while the AI processes the conversation.
4

Review the AI response

The results panel shows:
  • AI Response — The full text of what the AI said in response to the customer message.
  • Pass / Fail — An AI evaluator compares the response to your expected outcome and marks it as Pass (green) or Fail (red).
  • Evaluation Notes — A brief explanation of why it passed or failed.
Tests are text-based simulations, not real voice calls. They test the AI’s logic, knowledge, and decision-making. The actual voice quality and tone are not tested here — for that, use the Test AI page to make a real call.

Performance Stats

At the top of the AI Training page, you will see summary statistics:
  • Total Scenarios — How many scenarios you have created
  • Overall Pass Rate — The percentage of scenarios that passed their most recent test (e.g. “85% pass rate”)
  • Last Tested — When you last ran any test

Per-Scenario Tracking

Each scenario tracks its own history. Click into a scenario to see:
  • How many times it has been tested
  • The pass/fail result of each run
  • Whether performance has improved or declined over time

Deleting a Scenario

1

Find the scenario

Locate it in the scenario list.
2

Click Delete

Click the trash icon or Delete button on the scenario row.
3

Confirm deletion

A confirmation dialog asks if you are sure. Click Delete to confirm. The scenario and all its test history are permanently removed.

Best Practices

Start with 5 to 10 covering your most common call types: booking requests, pricing questions, emergency calls, after-hours enquiries, and returning customers. Add more as you encounter new situations in real calls.
This means the AI does not have enough information to handle that situation. Check your Knowledge Base and My Business page — is the relevant information there? Add it, then run the test again.
Run all your scenarios after any change to your AI configuration: greeting updates, Knowledge Base edits, new services, pricing changes, or additional instructions. A full test run takes under a minute.
If you have enabled Bilingual Mode in your AI settings, you can write scenarios in your secondary language. The AI will respond in that language, and the evaluator will assess the response accordingly.