How a session works
A learner picks a scenario and starts talking. The AI plays a person with a reason for behaving the way they do, not an obstacle to be defeated. It responds to what the learner actually says, so the conversation goes differently every time.
- Text or voice — speak your answer and hear the reply, on Pro plans
- Each scenario has a character, a situation and objectives the learner is trying to reach
- The AI follows behaviour rules written per scenario, so it stays in character
- At the end, the session is scored against a rubric with defined criteria
- Feedback quotes what the learner said and what to do differently
Scored against a rubric, not a vibe
Every scenario carries its own rubric — a handful of named criteria with written descriptions of what a 1, a 3 and a 5 look like. Scores are recorded per criterion and mapped to skills, so a pattern across many sessions becomes a real signal about where someone is strong and where they are not.
Two deliberate limits are worth stating plainly. One scenario is never treated as proof of a skill: a skill needs evidence from at least two different scenarios before it is reported as confirmed, so replaying the same conversation eight times does not manufacture a result. And a conversation too short to judge is refused rather than scored.
What managers see
- A readiness view per person, by skill, with the evidence behind each score
- Whether a score is AI-generated or has been reviewed and endorsed by a human
- A downloadable evaluation as PDF for appraisal or audit records
- Courses matched to the specific criteria a learner scored weakest on
On Pro plans, managers can review and endorse a session. Until a human signs one off, the report says so — a machine score is labelled as a machine score, because presenting it as a verified judgement is how an assessment tool loses trust.
Common questions
Is this replacing a trainer?
No. It is replacing the part where people learn on real customers. A trainer is limited by the hours in a day; practice is limited by nothing, and the trainer gets to spend their time on the people the reports show need it.
Do learners find it uncomfortable?
Less than the real conversation. The practice space is deliberately private — sessions are not shown to peers, and the leaderboard ranks effort such as scenarios attempted, never scores. Ranking people by assessed performance would push everyone toward the easy scenarios.
What languages does it work in?
Conversations run in English. The surrounding interface is available in English, Urdu and Arabic.
Can we write our own scenarios?
Yes. Administrators can create scenarios with their own character, situation, objectives and rubric — so the practice matches your products, your policies and your customers rather than a generic script.
Which plans include it?
Text practice is available from Grow. Voice practice, manager review and endorsement are on Pro and above. Each plan includes a monthly allowance of practice sessions per learner.
Try a scenario
It is difficult to judge this from a description — the point is how the conversation actually goes. Start a free trial and run one, or ask us to walk you through it.