Practice that talks back

Practise the conversations that matter

You will speak (or type) with an AI playing a real person from your day. When you finish, you get a scored docket against the same standards your team is marked on, plus coaching on exactly what to sharpen.

Before you start

Find somewhere you can speak out loud, or choose text mode on the briefing screen.
Your browser will ask for microphone access for voice practice.
Treat it like the real thing. The AI responds to how you actually handle it.

Step 1 of 3

What do you want to practise?

These come straight from your scenario library.

Step 2 of 3

Pick your difficulty

Step 3 of 3

Your briefing

Voice mode needs your microphone. Your browser will ask when the conversation starts.

0:00

Connecting

✓

Reading the transcript...

Admin

Edit the knowledge pack, tune rubrics, and build scenarios. Changes apply immediately for testing in this session. Use the export buttons to save files back into the package.

These slots are domain-agnostic. Reference sections replace things like menus, channels replace things like counter and drive-thru, exclusions are the "things we do not do" that the assessor never penalises a polite decline on.

Organisation name Reference sections (JSON array) Channels (JSON array) Policies (one per line) Exclusions (JSON array of {ask, decline}) Difficulty settings (JSON object) Manual voice overrides (one per line: key|label|voiceId)

Rubric pack

ID	Criterion	Weight	Applies to	What earns Done well

Auto-fail conditions (one per line, any of these caps the result at fail)

Upload CSV

Generate criteria with AI

Describe the conversation type and what good looks like

Give the builder a rough brief. It compiles a full scenario grounded in the knowledge pack, picks the right rubric pack and a voice, and you can test it immediately. Copy the JSON into scenarios.js to keep it.

Rough brief Extra rules (optional)