Retail customer service
Returns without a receipt, complaints at the counter, and the customer who wants the manager.
For employers
AI role-play training that gets your staff ready for real conversations.
Build spoken practice for the conversations your team actually has: complaints, check-ins, phone calls, visitors who speak another language. Staff rehearse out loud with an AI character and get scored against the goal you set.
Welcoming visitors
6 of 8 ready
Team workspace — now in pilot
Know who's ready before they face the public.
Every person practises the same scenarios and gets an honest score. You see who has passed, who needs another go and who hasn't started.
Guest services team · Welcoming visitors
6 of 8 ready
The team view is being built with pilot organisations.
How it works
From your situations to a team that's ready.
- Step 1Available now
Build your scenarios
Describe who the AI plays and what they want: an upset customer, a nervous patient, a visitor who's lost. Choose one-on-one or interpreter in the middle.
- Step 2Available now
Set the goal and what's assessed
Write the task card your staff see, the points they must cover, when the conversation is finished, and the time limit.
- Step 3Pilot
Invite your team
Staff get a private course and practise on their own time, in practice mode first, then assessed.
Welcoming visitors
6 of 8 ready
Step 4PilotSee who's ready
One view of who has passed, who needs more practice and who hasn't started, with every score.
Who it's for
Any team that talks to customers, patients or visitors.
If a conversation matters and your people only get one shot at it, they can rehearse it first.
Hospitality and front of house
Check-ins that go wrong, booking mix-ups, and guests with dietary or access needs.
Healthcare reception and intake
Checking in a worried patient, explaining wait times and calming a distressed family member.
Contact centres
Billing disputes, cancellations and hardship calls, rehearsed out loud before the first live call.
Volunteers for major events
Directions, lost property, ticket problems and access requests from international visitors, in English or their language.
Language services teams
Interpreters rehearse the interpreter-in-the-middle format in your domains before an assignment.
It also suits government and council front counters, banks and insurers, libraries, and any team where the first conversation sets the tone.
Any language, any goal
Your conversations, in the languages your public speaks.
One-on-one role-play
Your staff member plays themselves. The AI plays the customer, patient or caller.
Interpreter in the middle
Two AI people speak different languages. Your staff member interprets both ways.
English only, or English and one of 30 listed languages (or type another):
- English only
- Arabic
- Bangla
- Cantonese
- Dari
- Filipino
- French
- Greek
- Gujarati
- Haitian Creole
- Hindi
- Indonesian
- Italian
- Japanese
- Korean
- Malay
- Malayalam
- Mandarin
- Nepali
- Persian
- Portuguese
- Punjabi
- Russian
- Sinhala
- Spanish
- Tamil
- Telugu
- Thai
- Turkish
- Urdu
- Vietnamese
You decide what “done” looks like
It's finished when…
The customer accepts a store credit or a manager callback
It's finished when…
The patient's details are confirmed and an interpreter is booked
It's finished when…
The visitor knows how to reach the accessible entrance
The AI character works toward its own goal, and the session is scored on whether your staff member got the conversation there.
Assessment
Practise freely. Get assessed when it counts.
Scores are tied to the goal of the conversation, so getting halfway doesn't count as ready.
- Practice mode coaches every turn
- In a role-play, task-card points tick off as they're covered. When interpreting, each turn gets a quick verdict and a one-line tip, with up to three tries. Nothing is scored.
- Assessed sessions run like the real thing
- No hints. A score out of 100 at the end, with a target of 70.
- Unfinished means not ready
- If the conversation doesn't reach its goal, the score is scaled by how much of the task was covered and can't reach the target.
- Feedback people can act on
- Specific notes on what to fix next time. Scores are estimated from the transcript, so tone and pronunciation aren't judged.
Where things stand
What you can use today, and what we're building with pilot teams.
Available now
- Custom scenarios: one-on-one role-plays or interpreter in the middle
- AI voice characters that stay in role
- English only, or English and another language
- Task cards, a clear finish line and a time limit
- Practice mode with coaching, and assessed sessions
- A score and specific feedback for every assessed session
PilotTeam workspace
- Private courses only your team can see
- Inviting staff by email
- A team view of who's ready, with scores
- Assessment criteria set by your organisation
- Team reporting
Training interpreters or running NAATI test prep? See cohort access for training providers.
Questions from employers
What is AI role-play training for staff?+
Your staff member speaks out loud with an AI character who plays a customer, patient, caller or visitor. The character has its own goal and reacts to what your staff member says. Each session is timed, and assessed sessions are scored with specific feedback.
Can we build scenarios for our own situations?+
Yes, as part of a pilot. Each scenario sets who the AI plays, what they want, what your staff member plays, what's on their task card, when the conversation is finished, and the time limit. Book a pilot and we'll set them up with you.
Which languages can staff practise in?+
English only, or English and another language. The language picker lists 30 languages and staff can type any other. In a one-on-one role-play the AI character speaks the chosen language. In the interpreter format, one person speaks English, the other speaks the second language, and your staff member interprets both ways.
How is each session scored?+
Assessed role-plays are scored out of 100 across task completion, language, fluency, interaction, professionalism, with a target of 70. The score reflects whether the conversation reached its goal: an unfinished session is scaled by how much of the task was covered and can't reach the target. Scores are estimated from the transcript, so tone and pronunciation aren't judged.
Can we set our own assessment criteria?+
Today the role-play criteria are the same for every course, and you shape what counts through the goal, the task card and when the conversation is finished. Criteria set by your organisation are something we're working out with pilot teams.
Can I see which staff are ready?+
Not yet as a self-serve feature. Today each person sees their own scores and history. Inviting staff and a team view showing who's ready, who needs practice and who hasn't started are what we're building with pilot organisations. Get in touch if you'd like to be one.
What's the difference between practice and assessed sessions?+
In practice mode, role-play task-card points tick off as your staff member covers them, and in the interpreter format each turn gets a verdict and a one-line tip, with up to three tries. Practice sessions aren't scored. Assessed sessions run like the real thing and are scored at the end. Opening the transcript during an assessed session turns it into practice.
Does it work for volunteers at big events?+
Yes. Volunteers and guest-services staff can rehearse the questions international visitors ask most, such as directions, lost property, tickets and access needs, in English or with a visitor who speaks another language.
What do staff need?+
A web browser, a microphone and ideally headphones. Sessions run in the browser and a single session is capped at 30 minutes.
What does it cost for a team?+
There's no self-serve team plan yet. Tell us about your team, the conversations you want to train and the languages you need, and we'll work out a pilot with you.
Run a pilot with your team.
Tell us who you're training, the conversations that matter and the languages you need. We'll reply by email to set up a pilot.