AI Roleplay and Simulation Glossary for Training Teams

The basic words for understanding how a conversation with an AI becomes a practice exercise.

Roleplay (role play)

An exercise in which a person plays a work role opposite someone playing theirs, to practice a real conversation. With AI, the counterpart is virtual: it answers in text or by voice, and no colleague has to free up time. See the guide to AI sales roleplay.

Conversational simulation

An interactive scenario where the learner talks with a conversational agent playing a customer, a candidate, an employee or a manager. It reproduces the dynamics of a real exchange (follow-ups, silences, objections) and can be replayed as often as needed, with no risk to a real customer. AI-Coaching offers text and voice simulations. You can try a free demo, no account needed.

Virtual counterpart

The AI agent that plays the other side of the conversation: a rushed buyer, an unhappy customer, a hesitant candidate, an employee in difficulty. Its role, personality and reactions are set before the session. The learner has to convince, reassure or support it. See also persona.

Persona

A profile describing the character played by the virtual counterpart: job, context, personality, goals, constraints, knowledge of the topic and way of reacting. A precise persona makes practice believable and repeatable. In AI-Coaching, personas are configurable, so you can prepare the same situation against very different profiles.

Scenario

The outline of a simulation: context, learner role, counterpart role, goal and success criteria. A good scenario reflects a situation your teams really face. In AI-Coaching, a scenario can be enriched with your own documents (offers, procedures, sales pitches) through RAG.

Mission

The instruction given to the learner for a session: what to achieve in the conversation, such as booking a meeting, handling a complaint or conducting an interview. The mission makes a session measurable, since you know what had to be reached. It is prepared along with the scenario.

AI observer

An agent that follows the exchange without taking part. In AI-Coaching, it gives discreet real-time coaching during the session, then produces a detailed debrief against configurable criteria. It does not play the customer: it observes and advises, much like a coach sitting next to the learner. It complements a manager’s observation; it does not replace it.

Debrief report (bilan)

The summary produced at the end of a session: strengths, areas to improve, a score per criterion, key moments of the conversation. A useful report is precise, quotes passages of the exchange and leads to an action. In AI-Coaching, the AI observer writes it against your criteria.

Feedback

What the learner is told about what they did, during or after the exercise. Good feedback is factual, fast, focused on observable behaviors and comes with a way to improve. It can come from the AI observer, a trainer, a manager or peers.

Debriefing

The conversation that follows a simulation: the learner analyzes what they did and what they would do differently, alone, with a trainer or with a manager, starting from the report. This is where experience turns into learning. It can be one-to-one or in a group.

The notions that connect simulation to a training program and to skills tracking.

Work situation

The real context of an employee’s activity: tasks, tools, counterparts, constraints. It is the raw material of AFEST. An AI simulation lets people practice work situations that are rare, risky or hard to reproduce on the job, such as a very angry customer or a tense negotiation.

AFEST (action de formation en situation de travail)

A French training format recognized by the Labor Code, which uses work activity itself as the learning support. It requires a prior analysis of the activity, a designated trainer acting as a tutor, reflective phases separate from the work situation, and an assessment of what was learned. Check these conditions with your training fund (OPCO) before labeling a program as AFEST. See the guide to scaling work-based training with AI.

Competency framework

A document describing the skills expected for a job or role, with observable behaviors and proficiency levels. It is the shared basis for assessments, training plans and reviews. Its value comes from the precision of its criteria. The guide to the sales skills matrix shows how to build one for sales.

Skills mapping (skills matrix)

An overview of each person’s or team’s level, skill by skill, measured against a framework. It reveals individual and collective gaps and helps decide what to train, and for whom. AI-Coaching offers skills mapping and progress tracking, fed by simulation sessions.

Upskilling (montée en compétence)

A person’s progress toward the expected level on a skill. It shows in behaviors that are better and better mastered, session after session, more than in hours of training. Regular tracking makes it visible well before operational results move.

Learning path (parcours)

An ordered sequence of learning activities toward a goal: for example several simulations of increasing difficulty, along with debriefing time. In AI-Coaching, you build pathways from scenarios, to link each skills gap to specific exercises.

Playlist

A selection of scenarios grouped around a theme or need, offered to a learner or a group: for example “price objections” or “a new hire’s first weeks”. More flexible than a full curriculum, it suits recurring practice. AI-Coaching lets you organize your scenarios into pathways and playlists.

Onboarding

Bringing a new hire to the point where they are autonomous in the role. In sales and in support alike, simulation lets people practice before their first real conversations with customers. See the 30-60-90 day sales onboarding plan and the AI simulator for onboarding.

The technical terms you need when talking with IT, an LMS vendor or an AI provider.

LLM (large language model)

An AI model trained on very large volumes of text, able to understand an instruction and produce answers in natural language. It is the engine that makes the virtual counterpart talk. AI-Coaching lets you choose the LLM: Mistral, OpenAI, Azure, Gemini, deepseek or a self-hosted model.

RAG (retrieval-augmented generation)

Before answering, the system searches your documents for relevant passages and passes them to the LLM as context. The counterpart then reacts to your offer or procedures rather than to generic knowledge. The principle was described in a 2020 research paper. In AI-Coaching, scenarios are enriched with your documents through RAG.

Self-hosted model

A language model installed and run on infrastructure you control, instead of being called from an external provider. It meets data-control requirements, at the price of operations you must plan for. AI-Coaching runs in the cloud or on premises and supports a self-hosted LLM.

Speech synthesis (text-to-speech)

Technology that turns text into voice. It gives the virtual counterpart its voice. AI-Coaching offers ElevenLabs, OpenAI Realtime and Azure Speech voices for voice simulations.

Speech recognition (speech-to-text)

Technology that turns speech into text, so the system understands what the learner says. Its quality with accents, noise and job-specific vocabulary affects how smooth a voice exchange feels.

Latency

The delay between the end of what the learner says and the start of the counterpart’s answer. In spoken conversation, too much latency breaks the natural flow. It depends on the model, the voice engine, the network and the hosting.

Real time

Describes processing whose result arrives fast enough to keep up with a conversation. For a voice simulation, it means exchanging without a noticeable wait. For the AI observer, it means coaching during the session rather than afterward.

LMS (learning management system)

A platform that hosts courses, enrolls learners, tracks their progress and produces reports. It is often a learning department’s system of record. A simulation platform complements it for practice. See integrating AI simulations into an LMS.

LXP (learning experience platform)

A learner-centered platform that recommends and aggregates content from varied sources rather than imposing a catalog. It often coexists with an LMS, which keeps administrative tracking.

SCORM

Sharable Content Object Reference Model: a set of specifications created by ADL (Advanced Distributed Learning) that lets learning content be packaged and communicate with an LMS for launch and tracking, notably completion and score. Several editions exist, including SCORM 1.2 and SCORM 2004. AI-Coaching’s SCORM compatibility is available: see the page on AI simulation and SCORM.

xAPI (Experience API)

An ADL specification, also known as Tin Can, that describes learning experiences as statements (actor, verb, object) sent to a learning record store (LRS). More flexible than SCORM, it can track activities outside the LMS.

LTI

Learning Tools Interoperability: a 1EdTech standard that connects an external tool to a learning environment, so users open it from their LMS without signing in separately. It concerns launching a tool, not the format of the content.

Multi-profile

Access rights organized by role. In AI-Coaching, three profiles coexist: the operator administers the platform, the editor designs scenarios and the user practices.

Secure link sharing

Making an exercise available through an address sent to participants. In AI-Coaching, a link can be limited by an expiry date and a usage cap, so an exercise does not circulate without control.

The sales methods you find in AI sales roleplay scenarios.

Sales enablement

All the content, tools, training and practices that help sales teams sell better. Simulation is part of it: it lets reps absorb a pitch before testing it on a customer. See AI sales coaching.

Sales discovery

The phase where the seller questions the customer to understand their situation, stakes, constraints and decision criteria before proposing anything. It relies on open questions and on rephrasing. See sales discovery questions.

SPIN

A questioning method described by Neil Rackham, standing for Situation, Problem, Implication and Need-payoff. You first understand the context, bring out a problem, explore its consequences, then lead the customer to state the value of a solution themselves. It suits complex sales best.

BANT

A framework for qualifying an opportunity: Budget, Authority (who decides), Need and Timing. It helps decide whether a deal deserves time, but stays basic for complex sales involving several decision makers.

MEDDIC

A qualification method for complex deals: Metrics, Economic buyer, Decision criteria, Decision process, Identify pain and Champion. The MEDDICC variant adds Competition.

Objection

A hesitation or concern raised by the customer: price, timing, incumbent supplier, doubt. An objection is often a request for information or reassurance, handled by asking questions before answering. See the guide to handling sales objections.

Sales negotiation

The phase where seller and buyer adjust the terms of an agreement: price, scope, timing, commitment. A core principle: no concession without a trade. See sales negotiation with AI simulation.

BATNA

Best Alternative To a Negotiated Agreement: your best fallback if the negotiation fails. The concept comes from Roger Fisher and William Ury’s work on principled negotiation. Knowing your BATNA, and estimating the other party’s, helps set your limits before sitting down.

The notions that come up in support and customer service scenarios.

De-escalation

A set of techniques to lower tension with an angry counterpart: listen without interrupting, acknowledge the emotion, rephrase, then offer a concrete solution. It is learned through practice, and simulation with an unhappy virtual customer suits it well. See handling an unhappy customer with AI simulation.

Active listening

A way of listening that shows the other person they are understood: sustained attention, respected silences, open questions and rephrasing. It matters in sales, support and management. An observer can check whether the learner truly rephrases what was said.

CSAT

Customer Satisfaction Score: a measure of satisfaction after an interaction, from a question such as “How satisfied are you with this exchange?” rated on a short scale. It is usually reported as the percentage of positive answers. It measures one moment, not the overall relationship. See the AI simulator for customer support.

NPS

Net Promoter Score: a recommendation metric based on the question “How likely are you to recommend us?”, rated from 0 to 10. Scores of 9 and 10 are promoters, 0 to 6 detractors, 7 and 8 passives. The score is the percentage of promoters minus the percentage of detractors, a value between -100 and +100.

Benchmarks for judging whether a training program has an effect.

Kirkpatrick model

A four-level framework for evaluating training: reaction, learning, behavior and results. It runs from participants’ impressions to the effect on the organization. The higher you go, the more useful the measure, and the harder it is to isolate. See how to measure the impact of AI coaching.

Training ROI

The ratio of net gain to cost: (benefits minus costs) divided by costs, as a percentage. The difficulty is putting a figure on benefits and isolating training’s share. Any worked example should be presented as an assumption. See the sales training ROI calculator and the article on AI sales training ROI.

Time-to-productivity

The time it takes a new hire to reach the expected performance level. It is defined per role (first meeting, first sale, autonomy on tickets) and used to steer onboarding. No value is universal: compare your team with itself over time.

Scorecard (assessment grid)

A grid of scored criteria, with levels or weightings, to assess a conversation or a performance. In simulation, it is the basis of the debrief report. In sales, its criteria often cover discovery, argumentation, objection handling and the final commitment. See the assessment grid and sales framework.

Pages that go deeper into the terms of this glossary:

• AI sales roleplay: scenarios, personas and scoring grid.
• AI roleplay demo: try a session for free, no account.
• Sales skills matrix: framework, levels, assessment.
• LMS and SCORM integration: connect simulations to your learning platform.
• AI roleplay software: compare approaches.
• AI simulator for internal training and for trainers.