How role-play sessions feed the skills engine
Role-play is the most evidence-rich source of skill signals in GFoundry. The admin describes a scenario to the AI designer; the AI builds the role-play and assigns the skills it exercises. Each learner session then produces per-skill evidence based on the rubric scores.
Note: The role-play feature requires the role-play configuration to be enabled on your tenant. If you do not see a Role Play tab inside Learn Content, ask the GFoundry team to enable it.
How is a role-play created?
Role-play content is created inside a Learn Content item, through an AI designer that the admin works with in chat.
Go to Modules > Learn > Learn Content and open the content item.
In the Role Play block, select Create.
The Create a Role Play modal opens. Describe the scenario you want learners to practice in plain language. The placeholder reads "Describe the scenario you want learners to practice...".
The AI designer proposes a complete role-play: Persona, Language, Stages, Criteria, and Skills.
Iterate by replying to the assistant when something is off.
When satisfied, use Test role play to try the scenario as a learner would.
Use Publish to make the role-play available.
The disclaimer in the designer is correct: "AI suggestions are drafts. Test the role-play before publishing." Always run at least one test session before publishing.
How are skills assigned to a role-play?
Admins do not pick skills from a list. The AI designer derives the skill list from the scenario description and from the criteria it generates for the rubric. The result appears under Role Play Details in the Skills field, as read-only pills.
If you want the role-play to exercise different skills, change the scenario description so the assistant generates a different design. Specific guidance on the situation, the audience, and the desired behaviour tends to produce sharper skill assignments.
Where do I see the role-play structure?
Once published, the Role Play Details page shows the full template:
Persona: the character the AI plays during the simulation.
Language: the language of the conversation.
Stages: the phases the scenario goes through.
Criteria: the rubric items used to evaluate each learner session.
Skills: the skills the role-play exercises, derived by the AI.
The same fields are visible in the Test role play preview, under tabs Conversation and Evaluation, plus the side panels for Stages, Criteria, and Skills.
What happens when a learner runs a session?
Each learner session is evaluated against the rubric. For every criterion, the evaluator assigns a score. The skills linked to that criterion receive a contribution based on the score:
Score 2: full contribution.
Score 1: partial contribution.
Score 0: no contribution.
The contributions are aggregated into one evidence record per session per skill. The record is written once and never modified, even if the rubric or skill list changes later.
How is a skill validated for a user?
A role-play skill becomes validated for a user when two conditions are met across that user's sessions on role-plays that exercise the skill:
The user has at least 3 sessions with evidence for the skill.
The average evidence score across those sessions is 0.7 or above.
This rule is calculated at the moment the data is read, not stored. If the rule changes, the full session history is re-classified instantly, with no migration. The evidence records stay the same; only the validated flag flips.
Where do admins see the sessions and outcomes?
On the Role Play Details page, the Sessions list shows each session that learners have run. Each row links to a session detail view with the full conversation and the evaluation result, marked as Passed or Did not pass, along with feedback notes.
Per-user skill aggregates (which skills a person has accumulated evidence on, and which are validated) surface in Skills Mapping at the tenant level, and inside Gi Bot when a manager or admin asks about an employee's skills.
How does role-play compare to the other skill sources?
Role-play produces the cleanest skill signal of the four channels because the evidence is per session, scored from a rubric, and time-stamped. Learn completions, recognition endorsements, and evaluation ratings also feed the skills engine, but with coarser granularity.
A skill that should be highly validated for a learner usually shows up in at least two of these channels. Role-play alone is sufficient when a user has accumulated three or more sessions with strong scores; without that, the other channels need to compensate.