Role-Play Coach Agent: AI Skills Practice and Evaluation
Build role-play scenarios where learners practise a conversation with an AI persona by text or voice and receive a structured evaluation at the end.
What it does
The Role-Play Coach Agent lets you add a Role Play to a Learn Content: an interactive simulation where learners practise a conversation (feedback, sales, leadership, customer service) with an AI persona. When the conversation ends, the learner gets an evaluation against the criteria defined for the scenario.
Key capabilities
Scenarios built in a conversation: describe the situation to the Role-Play Coach Agent and it drafts the persona, the stages and the evaluation criteria for you.
A real dialogue, not scripted branches: the AI persona reacts to what the learner types or says, according to the mode configured for the role-play.
Progress through stages: the scenario is organised in stages, and the learner sees which stage they are in.
Structured evaluation at the end: a score per criterion with feedback, an overall score and a pass or fail result.
Text or voice
Each role-play uses one mode. In Text, the learner types and reads replies. In Voice, the learner talks into a microphone and hears the character respond in real time, with the pauses and interruptions of a spoken conversation.
The administrator chooses the mode in the designer chat. To change an existing role-play, open its details, select Edit and ask the designer to switch it to voice or text. The character's tone, pace, pauses and energy can also be refined there. Voice depends on availability for your organisation.
Use cases
Sales teams: discovery calls, objection handling, negotiation.
Managers and team leaders: feedback conversations, performance discussions, difficult announcements.
Customer service teams: escalations, complaint handling, de-escalation.
HR and L&D: interview practice, onboarding conversations, compliance scenarios.
How do I build and run a role-play?
A role-play lives inside a Learn Content item, as its Role Play block, and who can take it follows the audience of that Learn Content. The step-by-step guides are in the Learn collection:
Designing a Role Play with the AI Designer: describe the scenario and let the agent build the persona, milestones and evaluation criteria.
Testing and publishing a Role Play: try it as a learner, read the evaluation and publish.
The learner experience: practicing a Role Play: what learners see, how a session ends and how it is scored.
Reviewing Role Play sessions, evaluations and feedback: the Overview and Sessions tabs, and the conversation and evaluation of each session.
For what Role Play is and how it feeds the skills engine, see Role Play: what it is and why it matters.
Privacy and data handling
The Role-Play Coach Agent uses OpenAI to build scenarios, conduct conversations and generate evaluations. The provider receives the scenario, the persona and the conversation, including what the learner types or says. In voice mode, the learner's microphone audio is processed to conduct the spoken conversation. The learner's name and email are not added to the prompt. Conversations and evaluations are stored in GFoundry. The Data Processing Agreement, including the list of sub-processors, is available on request.
What it does not do
It does not replace human coaching. The agent gives consistent, scalable practice; people handle context, nuance and growth conversations.
It does not evaluate personality. The evaluation scores the criteria of the scenario, not the learner as a person.
It can make mistakes. The evaluation is generated by AI: read it as feedback on a practice session, not as a verdict on the learner. Test the role-play before publishing, and again after editing.
Tips for better practice
Make the scenario specific. "A customer asking for a 30% discount on a renewal" beats "a tough customer call".
Pair with a course. Use the Learning Designer Agent to build the underlying course, then add the role-play as the practice step.
Review and iterate. After a few sessions, read the conversations and refine the criteria based on what learners actually do.
