Using AI roleplay in Creator – Docebo Help & Support

Introduction

AI roleplay is a Creator feature that helps learners practice real-world conversations in a simulated environment, featuring a virtual avatar powered by synthetic voice technology.

This article explains how to add and use AI roleplay scenarios in a Creator lesson, including how to write effective prompts and provide supporting details so the AI can generate useful, realistic interactions. You can create AI roleplay scenarios for any type of conversation.

Learn how to create and manage Creator lessons and how to use the Creator tool.

Requirements and limitations

AI roleplay is only available if you have activated the corresponding card in the Artificial intelligence control panel.
While the learner experience for AI roleplay scenarios can be localized into all languages available for Creator lesson translations, the scenario configuration panel is available in English only, regardless of your platform language. Learn more on translating Creator lessons.

Creating AI roleplay scenarios

You can create AI roleplay scenarios in two ways:

When you create a new Creator lesson from the Navigation menu, select AI roleplay
From an existing Creator lesson, click on AI experience in the toolbar at the bottom of the Creator editor, then select AI roleplay.

In both cases, a screen will open where you can define your scenario, choosing between two options:

Enter a custom prompt: Describe the conversation scenario you want to create in the text field. Be specific about the situation, roles involved, and the goal of the conversation.
Select a featured template: Choose from pre-built templates designed for common business scenarios. Templates are organized by category.

You will be redirected to the configuration panel, with fields prefilled by the AI based on your prompt or on the selected template. You can review and adjust these fields as needed. Refer to the next chapter to learn more about the scenario configuration.

Please note! When using a template or a prompt, AI automatically fills configuration fields, except for the following, even if you mention them in the prompt:

Interaction mode (defaults to Video)
Time to complete (defaults to 5 minutes)
Number of attempts (defaults to 1)
Make scenario mandatory (defaults to disabled)
Scenario title (defaults to "AI roleplay scenario").

Configuring a scenario

When you enter a custom prompt or select a template, the scenario configuration panel opens. This panel is flexible and works for any type of conversation. You can configure several sections. Let’s go through them one by one.

Scenario overview

Define the context and objectives of the conversation scenario. Provide a clear description of the situation, set the learner’s goal, and add any optional context that will help the AI avatar respond realistically.

Scenario description

Write a short description of the situation and how the conversation begins. Keep it to one or two sentences.

Good example: An HR manager schedules a meeting with an employee to provide feedback on recent performance.
Poor example: A corporate board meeting. (Too vague: no roles, no flow.)
Poor example: A project manager at a large company that has recently gone through reorganizations and budget cuts meets with a senior engineer about missed deadlines. The project manager is juggling stakeholder expectations and the engineer feels frustrated about shifting priorities. (Too long: overloaded with background, goals get buried.)

Goal to achieve

Define the main result that the learner should reach in this scenario. Focus on what the learner must accomplish, not on what the avatar does.

Good example: The HR manager delivers constructive feedback, addresses concerns, and agrees on clear next steps.
Poor example: The employee listens carefully and accepts the manager’s suggestions. (Focuses on the avatar, not on the learner.)

Additional context

This field is optional. Use it to add any extra background that can enrich the scenario. Include details that don’t fit in the other fields but will help the AI generate more realistic answers.

Good example: The company has recently gone through a reorganization, creating uncertainty in the workplace.
Poor example: The office recently moved to a new building. (Not relevant to the scenario and does not affect the conversation.)

Conversation participants

Define who is involved in the conversation and their characteristics.

AI avatar

Click on Edit AI avatar to define who the AI avatar is, along with their background, motivation, and personality. This helps the AI generate realistic and consistent answers.

When configuring the avatar, you can also set its appearance. Use the Select avatar button to choose an image that represents the character in the scenario. The avatar’s look does not affect the conversation flow, but it makes the experience more engaging for learners. You can search avatars by name or filter them by appearance, age, and setting. Select the desired avatar by clicking on its card and press Confirm.

Please note! The avatar image is visible to learners in Video and Voice interaction modes. In Chat mode, the conversation is text-only and no avatar image is displayed.

Avatar name

Give the avatar a realistic name.

Example: Daniel

Avatar role

Specify the avatar’s professional role or position in the scenario.

Good example: Employee.
Poor example: Employee receiving feedback from their manager. (Too descriptive.)

Avatar profile

Describe the avatar’s background and motivation in this scenario.

Good example: The employee is worried about their performance and hopes to understand how to improve.
Poor example: The employee works at a multinational company with 200,000 staff and complex reporting lines. (Irrelevant detail, not shaping the scenario.)

Avatar personality

Choose one of the predefined options (Agreeable, Vague, Skeptical) or define a custom personality by combining 2–3 traits.

Good example: Open to feedback, but defensive when criticized too directly.
Poor example: Very kind, collaborative, skeptical, stressed, vague, friendly, approachable, nervous. (Too many traits, confusing.)

Learner

Click on Edit learner to define the role that the learner will play in the scenario. This ensures the AI understands who the learner is and how to respond to them.

Learner role

Describe the learner’s role in the scenario. Keep it short and role-focused, without adding background story.

Good example: HR manager giving feedback to an employee.
Poor example: Manager in a company. (Too generic.)

Interaction mode

Select how learners will interact with the AI avatar during the scenario.

Video: Learners interact with an animated AI character that speaks and moves on screen. This mode provides the most immersive experience.
Voice: Learners talk with the AI character through voice while a static avatar image appears on screen.
Chat: Learners exchange written messages with the AI character through text only.

Evaluation

In the scenario settings panel, the Evaluation section lets you define what the AI provides feedback on. By default, every scenario uses a standard rubric based on general communication criteria. You can replace it with a custom rubric that reflects your own coaching standards.

A custom rubric lets you define up to five criteria, each describing what "Good" and "Developing" looks like in the context of your scenario. When a learner completes the simulation, the AI uses your criteria to structure the feedback in their report.

Configuring a rubric

In the scenario settings panel, scroll to the Evaluation section. The Rubric item shows Default until you customize it.

To edit the rubric, click the pencil icon next to Rubric. The Edit rubric modal opens with a default rubric pre-loaded as a starting point.

Each criterion has four fields:

Criterion title: A short, specific name for the skill or behavior being assessed (for example, "Active listening" or "Objection handling").
Description: One or two sentences explaining what this criterion covers. Write it as if you are briefing a new reviewer.
What Good looks like: Concrete, observable behaviors that indicate strong performance. You can include example phrases, actions, or required elements.
What Developing looks like: Concrete, observable behaviors that indicate areas for improvement.

You can:

Add a criterion by clicking Add criterion at the bottom of the list (up to 5 total).
Delete a criterion by clicking the ellipsis icon on any criterion card and selecting Delete. At least one criterion must remain.

When your criteria are ready, click Save changes. The rubric is saved as part of the scenario configuration and becomes active when you publish the scenario.

The AI generates feedback based on the conversation transcript only. When writing criteria, focus on what can be observed in what was said — not on vocal delivery, body language, or other aspects that cannot be inferred from text.

If no custom rubric is configured, the scenario uses the default rubric and report structure.

Learn how to write effective custom rubrics.

Generating criteria with AI

Instead of writing criteria manually, you can ask AI to generate them based on a description of what you want to evaluate.

In the Edit rubric panel, click Generate criteria.

In the text field, describe what you want the rubric to evaluate. For example: "Evaluate whether the learner handles pricing objections by acknowledging the concern before presenting value."

Optionally, you can upload up to three reference documents (PDF or DOCX, up to 100 MB each) such as a sales playbook, coaching guide, or competency framework. The AI uses them as additional context for generation. Images contained in the documents are not processed.

When ready, click Continue. If the AI needs more context, it may ask one to three clarifying questions before generating. Select one of the suggested answers or type your own, then click Generate.

The generated criteria appear in the rubric form. Review and edit them as needed, then click Save changes.

Please note!

The scenario configuration (description, goal, avatar role, and learner role) is automatically used as context for generation. Make sure these fields are filled in before generating criteria.

Generated criteria replace any existing criteria in the rubric. If you already have criteria defined, a confirmation prompt appears before overwriting.

If the AI cannot generate criteria from your input — for example if the description is unrelated to a conversation scenario — an error message explains why. You can revise your description and try again, or write criteria manually.

Advanced settings

Press the Settings button at the bottom of the panel to open an additional configuration window. Here you can define the following options:

Time to complete: Set a time limit for the scenario. This option is only available in video and voice mode and can range from 5 to 20 minutes.
Number of attempts: Define how many times the learner can attempt the scenario.
Make scenario mandatory: If you select this option, the learner cannot complete the Creator lesson without performing the scenario.
Introduction card: Set the information shown to learners before the scenario starts. Enter a title; the description is automatically generated by AI once the scenario is created. You can edit the title at any time by clicking on it directly on the introduction card.

Once you have set up your scenario, press the Save configuration button at the bottom of the panel. To see the learner experience, preview the Creator lesson.

Configuring a legacy sales scenario

Please note! It is no longer possible to create new sales scenarios. Sales scenarios now use a configuration panel aligned with the standard scenario panel, so scenario setup follows a more consistent structure across scenario types. Legacy sales scenarios remain available and continue to work, but they do not receive new functionality. If you want to migrate an existing legacy sales scenario to the updated configuration, get in touch with your Customer success manager or contact us through Docebo Community (opens in a new tab).

The following section documents the configuration of existing legacy sales scenarios.

When you select a sales template, the sales scenario configuration panel opens. This panel is optimized for sales conversations with dedicated fields for customer profiles and company information. You can configure several sections. Let’s go through them one by one.

Interaction mode

Select how learners will interact with the AI character during the scenario.

Video: Learners interact with an animated AI character that speaks and moves on screen. This mode provides the most immersive experience.
Voice: Learners talk with the AI character through voice while a static avatar image appears on screen.
Chat: Learners exchange written messages with the AI character through text only.

Scenario

Select one of the available options:

Understanding needs: The learner practices how to ask the right questions and uncover the customer’s goals and challenges
Addressing concerns: The learner responds to customer objections and builds trust through effective communication. For this scenario, you need to add at least one concern in the corresponding text field, but we suggest including two or three for the best results. Ensure each concern is specific, realistic, and clearly separated from the others. You can include different types of concerns to cover varied situations.
Offering solutions: The learner presents a product or service as a tailored solution, focusing on value and benefits for the customer.

As an option, you can add context to your scenario. Click on the Add button to add:

Customer knowledge: Describe what the customer already knows before the simulation begins. This may include information from previous conversations with the learner, such as having seen a demo or reviewed pricing. It helps the AI interpret the customer as being partially informed and adjust their responses accordingly.
Scenario details: Add contextual information that helps the AI align the customer’s responses with the scenario. This field is flexible and can include anything relevant to the situation—like company priorities, internal blockers, or specific product considerations the learner should be aware of.

Customer profile

In this section, you can select one of the predefined profiles, optionally edit it, or create a new customer profile from scratch. Click on the ellipsis icon in the desired profile’s row to edit or delete the profile, or click on New customer profile to create one. You will need to enter a customer name and description. Press Save changes when you’re done.

AI character

Select one of the available characters or create a new one. Click on the ellipsis icon in the desired character’s row to edit or delete the character, or click on New character to create a new one.

Click on Select avatar to browse the avatar library. You can search avatars by name or filter them by appearance, age, and setting. Select the desired avatar by clicking on its card and press Confirm.

You can also define the character’s name, role, and personality:

Approachable: Open to dialogue but resistant to change. This customer is friendly and collaborative, yet hesitant to shift from current solutions without strong justification.
Vague: Evasive and hard to pin down. This personality avoids commitment and gives unclear or indirect answers, requiring the learner to ask targeted, clarifying questions.
Skeptical: Questions assumptions and pushes back. This type of customer challenges the learner, raising objections or doubts that must be addressed with confidence and persuasive reasoning.

Press Save changes when you’re done.

My company

In this section, you can use the predefined company or create your own company profile. Click on the ellipsis icon in the desired company’s row to edit or delete the company, or click on New company to create one. You will need to enter a company name and description. Press Save changes when you’re done.

Advanced settings

To configure additional options, refer to the Advanced settings section of the AI scenario configuration.

Click on Save settings to apply your changes.

Once you have set up your scenario, press the Save configuration button at the bottom of the panel. To see the learner experience, preview the Creator lesson.

AI-generated reports

At the end of each simulation, the AI generates a report based on the learner’s interaction. The report highlights key strengths, areas for improvement, and includes personalized suggestions to support skill development.

Here’s how it works in practice:

If the scenario is in voice mode, the AI transcribes the learner’s spoken input in real time, processes the transcription, generates a response, and converts it to spoken output for the conversation to continue
If the scenario is in chat mode, the interaction happens entirely via written text
At the end of the scenario, the full transcript of the exchange—spoken or written—is analyzed by the AI. A series of prompts is then used to generate a structured report.

Custom rubric report

When a custom rubric is configured for the scenario, the report is structured around your criteria instead of the default communication dimensions. For each criterion, the report includes:

The outcome: Good or Developing
The observable behaviors that drove the AI's feedback
Direct quotes from the conversation as supporting evidence.

This allows learners to understand how they performed against the specific standards you defined, and to identify what to work on based on concrete moments from their conversation.

Default rubric report

If the default rubric is maintained, the report provides feedback on general communication dimensions, including:

An overall feedback with strengths and areas for improvement
Tone of voice (intended as overall communication style inferred by the scenario's transcript, whether spoken or written), with attention to:
- Clarity – how clearly ideas are expressed
- Confidence – how assured and professional the communication feels
- Energy – the overall engagement and dynamism in the learner’s approach
Key quotes, where strong and weak moments in the learner’s performance are identified and highlighted.

The report is available immediately after the scenario and can be accessed again at any time by the learner from the course player. Learners can review it as part of their self-reflection. At this stage, the report and the corresponding transcript or recording are visible only to the learner and are not shared with managers or administrators.

After reviewing their report, learners can provide feedback on their experience with the scenario directly from the scenario summary page. This feedback helps Docebo improve the AI roleplay feature. Please note that learner feedback is anonymous, is not shared with platform administrators, and is used internally by Docebo for product improvement purposes only.

Learn more about the models used by AI roleplay.

Best practices

When creating a scenario from a custom prompt:

Be specific and structured: Include the situation, the roles your learner and the AI should play, and the goal clearly. Example: "The learner plays the role of an IT support analyst assisting a UX designer with a forgotten password, focusing on clear guidance, reassurance, and a professional, patient tone throughout the reset process."
Avoid overly long prompts: Keep your prompt focused on the essential elements. The AI will fill in most configuration fields, so you don't need to describe every detail.
Focus on the conversation goal: Clearly state what the learner should achieve by the end of the scenario.
Remember what AI won't extract: The AI cannot automatically set interaction mode, time to complete, number of attempts, whether the scenario is mandatory, or its title. You will need to configure these manually in the advanced settings.

When refining an AI roleplay scenario, keep these guidelines in mind to get the best results:

Keep inputs concise but complete: limit each field to 2–3 key points. Avoid long background stories that dilute the conversation.
Provide relevant context only: include details that influence the conversation. Leave out information that doesn’t affect the flow.
Choose the right persona:
- Skeptical: to practice handling objections.
- Agreeable: to practice collaboration.
- Vague: to practice clarifying unclear input.
Remember the purpose: AI roleplay is designed to help measure learners’ conversational capabilities and provide actionable feedback for improvement, rather than generate final scripts.

Best practices for scenario configuration

Fill out all fields thoughtfully: Use each field for its intended purpose. Don’t put all details in the Scenario or Additional context fields. Don’t repeat or overload fields.
Keep context short: The Additional context field is optional and should not contain any information that belongs in the other fields. Use it only if you need to add extra details, and limit it to 1–2 short sentences.
Avoid contradictions: make sure the information in the various fields is consistent.
Reference roles consistently: always use the same role name for the learner and the avatar across all fields (for example, if the learner is “employee” in one field, don’t call them “UX designer” in another).
Limit custom personalities: when defining a custom personality, use only 2–3 traits. Long lists of adjectives confuse the AI and make the avatar inconsistent.