Learning a skill and using it when it actually matters are two very different things. A course can walk someone through the right framework. A quiz can confirm they understood it. But the moment that framework needs to show up in a live conversation, under pressure, with a real person on the other side, is a different situation entirely.
This is not a new problem in L&D. Teams have tried to solve it in different ways. Manager led practice sessions, Peer to peer roleplay, and Shadow calls. Each of these works to a degree, but none of them scale. Practice ends up being rationed, feedback ends up being generic, and whether someone actually improved at a specific skill over time remains largely a matter of opinion.
There has not been a clean answer to this. Until now. That is what Lyearn AI Roleplay is built for.
What We Built and Why
Lyearn AI Roleplay gives learners a structured environment to practice real scenarios, get feedback tied to specific moments in their own conversation, and produce measurable skill data that connects directly to everything else happening in their development path inside Lyearn.
Three problems drove the decision to build this.
- Practice is rationed: Manager led sessions happen weekly at best. By the time a learner gets feedback on something they fumbled, they have repeated the same mistake several times over.
- Feedback is generic: Telling someone they sounded hesitant is not the same as showing them the exact moment in the conversation where it happened and what it cost them.
- Improvement is unmeasurable: There is no reliable way to answer whether someone is better at a specific skill today than they were five weeks ago, without relying on a subjective opinion.
How It Works
Lyearn AI Roleplay operates across three layers: creation, practice, and feedback.
Creation
Building an AI Roleplay in Lyearn is straightforward. The flow is sequential and takes only a few minutes from start to finish.
Scenario and Objective: This is where the roleplay begins. You can define what the conversation is about, what the learner is expected to achieve, and the context they are walking into. Relevant documents or reference materials can be added at this stage to give the AI more to work with. Pre-defined templates are also available for teams that want a faster starting point.
Avatar and Learner Role: Define who the learner is in the conversation and who they will be speaking with. The AI avatar comes with a name, job title, relationship to the learner, and personality cues that shape how it behaves during the session. Lyearn can generate the full avatar profile automatically, or you can build it manually.
Criteria and Rubric: This is what the learner gets scored against. Define the criteria, and each one maps directly to a Skill, and where relevant a Mindset, inside Lyearn's ontology. Objections the avatar should raise during the conversation can also be added here as an optional layer.
Delivery Configuration: The final step covers how the AI Roleplay is delivered and completed. You can set the conversation time limit, passing score, points, and whether a certificate is issued on completion. A cover image can also be configured here.
Practice
This is what the learner experiences when they enter an AI Roleplay session.
- Learners access AI Roleplay through their assignment notification or directly from the Library.
- Before starting, they see the full context: the scenario, their role, the AI avatar, and the exact criteria they will be scored against. Nothing is hidden before the conversation begins.
- The conversation runs live in voice or text, with sub-second response latency as the target.
- The AI avatar holds its persona throughout, pushes back based on what the learner actually says, and raises the difficulty if the learner is moving through too comfortably.
- Sessions end when the learner chooses to wrap up or when the configured time limit is hit.
- Learners can retry as many times as they want, with each attempt tracked and stored separately
Feedback
This is where Lyearn AI Roleplay produces something most practice tools do not.
- Every rubric criterion is scored with feedback grounded in verbatim quotes pulled directly from the learner's own conversation.
- The scoring is LLM-driven, using the rubric and the full conversation transcript as inputs.
- The reasoning behind every score is fully visible: which quote, which criterion, and what the model concluded. Learners can interrogate the feedback rather than just accept a number.
- A speech analytics layer tracks pacing, filler word frequency, clarity, and pause confidence across the session.
- Each attempt is stored separately, giving both the learner and the manager a visible improvement curve over time.
The Feedback Is the Product
Most tools give learners a score. Lyearn AI Roleplay shows learners why they got that score, with the exact words from their own conversation as the evidence.
Generic feedback does not change behavior. Knowing you scored a three out of five on objection handling tells a learner very little. Seeing the specific moment in the transcript where they conceded too early, with the model's reasoning attached, gives them something they can actually work with in the next attempt.
The speech analytics layer adds a second dimension. Pacing, filler word frequency, clarity of response, and pause confidence are tracked across every session. These are not soft observations. They are data points that accumulate across attempts and show a clear improvement curve, or the absence of one.
Each attempt is stored separately, so both the learner and the manager have a visible trajectory over time, not just a single snapshot.
x
Practice That Connects to Everything Else
Most practice tools produce a score. That score sits inside the tool, visible to no one, connected to nothing. When the session ends, the data ends with it.
Lyearn AI Roleplay works differently. Every attempt produces skill signal that feeds directly into Lyearn's MGSA framework, the same framework that every other activity inside Lyearn contributes to. Courses, quizzes, tasks, surveys and now AI Roleplay attempts, all producing data on the same scale, mapped to the same Skills and Mindsets.
What this means in practice is that a learner's AI Roleplay attempts do not exist in isolation. They sit alongside everything else in their development path and contribute to a longitudinal view of how their skills are actually moving over time. Not just whether they practiced, but whether the practice is producing measurable growth on the skills that matter.
For L&D teams, this closes a gap that has always existed. Practice has historically been the one part of the learning journey that produced no data. With Lyearn AI Roleplay, it does.
What This Changes for L&D
With Lyearn AI Roleplay, L&D leaders can now answer a question that has always been difficult to answer with any precision: is this person genuinely better at this skill than they were six weeks ago?
The answer no longer depends on a manager's recollection or a completion certificate. It comes from attempt data, score progression, verbatim feedback, and speech analytics, all sitting inside the same system as the rest of the learner's activity.
Practice is no longer something that happens outside the system and produces no data. It is part of the learning record, connected to skill signals, and visible to the people responsible for developing their teams.
Lyearn AI Roleplay is available now. You can create your first scenario from Learn > Library > Create Roleplay inside your Lyearn account.
