Mirror Roleplay FAQs

An AI Roleplay is a conversational simulation within Whatfix Mirror where an AI Persona plays a role, such as a customer, manager, or stakeholder, and responds dynamically to what the learner says. Learners use an AI Roleplay to practice interpersonal and decision-making skills in a realistic, risk-free environment before they apply those skills in real-world situations.

For more information, see Introduction to Mirror Roleplay.

Mirror Roleplay is available as part of an add-on package to Whatfix Mirror. To enable this capability for your account, contact your Customer Success Manager.

After Whatfix enables Mirror Roleplay, you create and manage AI Roleplays from the AI Roleplay section of the Whatfix Mirror Dashboard. Only users with the Account Manager or Content Manager role can access the Whatfix Mirror Dashboard.

For more information, see Access and Manage the Mirror Dashboard.

A Workflow enables your learners to practice the steps of a task in a replica of your application, and an AI Roleplay adds a conversational layer with an AI Persona to that practice. For example, a learner follows the steps to resolve a support ticket while the AI Persona plays a frustrated customer who wants a quick resolution.

You can create an AI Roleplay separately and then attach it to a Workflow, or attach a Workflow to an existing AI Roleplay. Once attached, the Workflow and the AI Roleplay move together across all stages of the Whatfix Mirror Dashboard.

Note:

You can attach an AI Roleplay to a Workflow, or a Workflow to an AI Roleplay, only in the Draft stage.

For more information, see Attach a Roleplay in Mirror.

An AI Roleplay consists of the following components. The examples are for a Roleplay in which a support agent handles a customer whose order has been delayed:

• Scenario: Define the situation, the AI Persona's emotional baseline and triggers, and how the conversation resolves. For example, the customer calls about an order that has been delayed twice. The customer starts frustrated, becomes more irritated when the learner gives vague answers, and calms down once the learner acknowledges the issue and offers a clear resolution.

• Persona: Define the AI Persona's name, job title, gender, language, voice, and behavioral traits. For example, Jane Doe, Operations Manager, who speaks English with a British accent and is impatient but open to reassurance.

• Evaluation: Define the parameters used to assess learner performance, including competencies and score weights. For example, an Empathy parameter scored on a 0–10 scale and mapped to the Customer Communication competency, alongside parameters for tone and accuracy.

• Exit Conditions: Define the time limit, repetition limit, and custom conditions that end the conversation. For example, a 10-minute time limit, a repetition limit of 3, and a custom condition that ends the session when the customer asks to speak to a manager.

• Learner Brief: Provide the learner with the context, objective, and role expectations before the conversation starts. For example, "You are a support agent handling a customer frustrated about a delayed order. Listen actively, acknowledge the issue, and provide a clear resolution to de-escalate the situation."

For more information, see Create an AI Roleplay.

Mirror supports the following Roleplay conversation types:

• Cold Calling

• Negotiation

• Discovery

• Renewal Management

• Product Pitch

• Mutual Discovery

• Competitive Difference

• Inbound Call

• Other

• Discovery with Evaluation

Each type focuses on a specific stage of a customer interaction or communication objective.

For more information, see Understand the Types of Roleplay Conversations.

Yes. You can design a Roleplay for organization-specific or role-specific interactions that do not align with the predefined conversation types. For example, a manager who conducts a quarterly performance review with an employee.

To discuss a custom Roleplay scenario for your training requirement, contact your Customer Success Manager.

Info:

You can change the Conversation type of a Roleplay after you create it, from the Conversation starters section of the Persona tab.

For more information, see Understand the Types of Roleplay Conversations.

Creating an AI Roleplay depends on whether you want Whatfix to generate the initial setup or you want to configure it yourself:

• Using the AI Roleplay Builder: Describe the scenario in a few lines, or upload a call transcript or reference document, and Whatfix generates the AI Persona, learner role and objective, evaluation parameters, modality, and difficulty level for you to review and modify. For more information, see Create an AI Roleplay using the AI Roleplay Builder.

• Manually: Define and configure every component of the Roleplay yourself, including the Scenario, Persona, Evaluation, Exit conditions and Learner brief. For more information, see Create an AI Roleplay Manually.

Yes. In the AI Roleplay Builder, click the Attachments icon to upload a call transcript or reference document instead of typing a description. Whatfix supports .doc, .docx and .txt files up to 5 MB and does not process images included in an uploaded file.

Avoid uploading Persönlich identifizierbare Informationen (PII) or sensitive information, such as names, email addresses, phone numbers, account numbers, or financial information. Whatfix discards an uploaded transcript after it processes the transcript and does not retain it.

Info:

Whatfix anonymizes all information it extracts from a transcript. Real customer names, email addresses, phone numbers, account numbers, company names, and internal process names never appear in a generated Roleplay. Instead, Whatfix substitutes them with fictional replacements throughout.

Note:

Whatfix cannot process a transcript with fewer than five interactions, or a transcript it cannot read. Upload a longer transcript, or describe the scenario in the text box instead.

For more information, see Create an AI Roleplay using the AI Roleplay Builder.

Describe the Roleplay in a few clear sentences so that the AI Roleplay Builder has enough information to generate an accurate setup. A vague description produces a generic Persona and Evaluation parameters that you need to rework later. Include the following elements:

• Situation: What has happened and why the conversation takes place. For example, an order that has been delayed twice.

• Learner role: Who the learner plays in the conversation. For example, a support agent.

• AI Persona role: Who the learner speaks to. For example, a frustrated customer.

• Desired outcome: What the learner should achieve by the end of the conversation. For example, de-escalate the conversation and provide an appropriate resolution.

Note:

Your description must contain at least 10 words and no more than 2,000 characters. To get started, select a template from the AI Roleplay Builder and adapt it to your scenario.

For more information, see Create an AI Roleplay using the AI Roleplay Builder and Best Practices for Mirror with Roleplay.

The difficulty level defines the level of challenge the learner faces during the Roleplay. The AI Roleplay Builder generates a difficulty level, which you can change when you review the generated setup.

Match the difficulty level to the learner's experience:

• Easy: Use for new learners and straightforward scenarios with fewer challenges.

• Medium: Use for learners who can apply the required skills more independently.

• Hard: Use for advanced learners who need to apply the required skills in complex or emotionally charged situations.

For example, select Easy with a confused but cooperative customer for a new hire in their first week, and Hard with an impatient customer for an experienced agent.

For more information, see Create an AI Roleplay using the AI Roleplay Builder and Best Practices for Mirror with Roleplay.

The Emotional state defines how the AI Persona feels during the interaction and influences how it responds to the learner. Select an emotional state that creates the challenge the learner needs to practice. For example, if the learner is practicing de-escalation, select a frustrated emotional state rather than a calm or cheerful one.

You set the Emotional state when you review the setup that the AI Roleplay Builder generates. If none of the available emotional states fit your scenario, click + Add custom to add a new one.

Tip:

Keep the Emotional state consistent with the Persona's Behavioural description and the Learner brief. A behavioural description that contradicts the emotional state can cause the AI Persona to respond unpredictably, and a Learner brief that contradicts it prepares the learner for the wrong conversation.

For more information, see Create an AI Roleplay using the AI Roleplay Builder and Best Practices for Mirror with Roleplay.

The interaction format, or Modality, depends on how you want learners to communicate with the AI Persona:

• Voice: Learners interact with the AI Persona using voice.

• Chat: Learners interact with the AI Persona using text.

Whatfix selects Audio roleplay as the default Modality when you create a Roleplay manually. For more information, see Create an AI Roleplay Manually.

Yes. On the Persona tab, in the Conversation starters section, select who starts the conversation, and then enter the Opening message and Closing message:

• AI persona: The AI Persona starts the conversation. Use this when the AI Persona initiates contact, such as in an inbound call.

• Learner: The learner starts the conversation. Use this when the AI Persona represents the person initiating contact, such as a customer in an inbound call.

  

For more information, see Create an AI Roleplay Manually and Best Practices for Mirror with Roleplay.

Creating or changing a Persona depends on how you create the Roleplay, or whether the Roleplay already exists:

• Using the AI Roleplay Builder: Whatfix generates a Persona from your scenario description. Click the generated Persona to customize it, click + Add custom to create a new one, or select an existing Persona. For more information, see Configure a Persona for an AI Roleplay.

• Manually: On the Persona tab, click Create persona and define the Persona's name, job title, gender, language, voice, and behavioural description, or click Choose persona to reuse an existing one. For more information, see Configure a Persona for an AI Roleplay.

• For an existing Roleplay: On the Persona tab, click Change persona, and then click + Add a new person to create a Persona or select an existing Persona from the list. For more information, see Edit a Roleplay in Mirror.

Mirror Roleplay supports 406 voices across 26 languages, including Bulgarian, Chinese, Dutch, English, French, German, Hindi, Italian, Japanese, Korean, Portuguese, Russian, Spanish, and more. Many languages offer multiple accent variants, such as American, British, Indian, and Australian for English, so you can align the Persona's voice with your training audience.

For more information, see Configure a Persona for an AI Roleplay.

Yes. Persona attributes apply to all Roleplays that use the same Persona. If you modify a Persona that other Roleplays use, review those Roleplays to confirm that the changes suit their scenarios.

To change the AI Persona for one Roleplay only, click Change persona on the Persona tab of that Roleplay and create a new Persona.

For more information, see Configure a Persona for an AI Roleplay.

Evaluation parameters define what a learner must demonstrate during a Roleplay and how their performance is scored. You can add up to 10 evaluation parameters to a single Roleplay.

For example, in a Roleplay where a support agent handles a customer whose order has been delayed, you can add Empathy, Tone, Accuracy, and Resolution as evaluation parameters. You can then score Empathy on a 0–10 scale and map it to the Customer Communication competency.

Use the following steps to add an evaluation parameter to an AI Roleplay:

1. Click the Evaluation tab.

   

2. Click + Add parameter.

   

3. Enter a name for the evaluation parameter. For example, Empathy. 

4. Click the Competency dropdown to open the competency selection list.  

   

5. Select the required competencies for the evaluation parameter by using the corresponding checkboxes.

   

6. Enter the Scoring guidance for the evaluation parameter. Define the level of learner performance that corresponds to each score range. For example, describe what scores of 0–2, 3–5, 6–8, and 9–10 indicate based on the learner’s performance.

   

7. In the Reference examples section, under Good response, enter examples of a good response.
   Example: “I understand how frustrating this must be. I’ll look into what happened with your order and help get this resolved.”

   

8. In the Reference examples section, under Poor response, enter examples of a bad response.
   Example: “You’ll need to contact billing. There’s nothing I can do about it.”

   

   Info:

   To add additional evaluation parameters, click + Add parameter and repeat the steps above.

   

After you add the required evaluation parameters, save the Roleplay.

For more information, see Define Evaluation Parameters.

Whatfix supports the following evaluation parameter types:

• Rating: Evaluates performance using a predefined numerical scale, such as 0 to 10.

• Text: Evaluates performance using written feedback or qualitative observations.

• Single Choice: Evaluates performance by selecting one option from a predefined list.

• Multiple Choice: Evaluates performance by selecting one or more options from a predefined list.

For more information, see Define Evaluation Parameters.

Competencies are the skills your organization tracks across Roleplays. After you add a competency, you can map it to an evaluation parameter in any AI Roleplay, so learner scores connect to that skill and you can compare performance across related Roleplays.

For example, you can add a Customer Communication competency and map it to the Empathy parameter of a Roleplay in which a support agent handles a customer whose order has been delayed. If you also map it to the Empathy parameter of a refund Roleplay, you can compare learner performance on that skill across both Roleplays.

Use the following steps to add a competency:

1. On the Whatfix Mirror Dashboard, click Settings.

   

2. Click AI Roleplay.

   

3. Click Competencies.

   

4. Click + Add competency.

   

5. Enter a name for the competency.

   

6. Enter a description for the competency.  

   

7. Click Add.

Exit conditions control when and under what circumstances the AI Persona ends a Roleplay session. Whatfix supports the following end conditions:

• Time limit: Ends the session after a configured duration, up to 15 minutes.

• Repetition limit: Ends the session after the AI Persona repeats the same question a set number of times, up to 5 repetitions.

• Custom conditions: Ends the session when a rule you define is met, such as when the customer asks to speak to a manager.

For more information, see Specify Exit Conditions.

You can configure a Roleplay session to last up to 15 minutes. Set the time limit from the Exit conditions tab of the AI Roleplay.

For more information, see Create an AI Roleplay Manually.

Whatfix evaluates a Roleplay attempt only up to the time limit you configure. Anything the learner says after the time limit is not considered for evaluation.

The attempt is still scored against the evaluation parameters and appears in Roleplay Analytics, as long as it lasted at least one minute.

A Learner brief gives a learner the context, objective, and role expectations for a Roleplay before the conversation starts. You can add a Learner brief of up to 500 characters, and Whatfix generates one automatically when you use the AI Roleplay Builder.

For more information, see Create an AI Roleplay.

Whatfix does not save changes to an AI Roleplay automatically, and the preview reflects the most recently saved version of the AI Roleplay. Click Save each time you make a change, and then preview the Roleplay again.

For more information, see Preview a Roleplay in Mirror.

Whatfix provides two ways to preview a Roleplay:

• Preview AI Roleplay: Tests the conversation with the AI Persona on its own. Access the AI Roleplay, click Preview AI Roleplay, and then click Start Session.

  

• Preview with Roleplay: Tests the complete experience your learners encounter when they access the Roleplay through the Workflow. Click the Workflow with the Roleplay enabled, and then click Preview with Roleplay.

  

Test the intended conversation first, and then deliberately deviate from it. For example, give brief answers, introduce off-topic comments, or avoid answering the Persona's question, to confirm that the Persona stays consistent and that the Exit conditions end the session as expected.

For more information, see Preview a Roleplay in Mirror and Best Practices for Mirror with Roleplay.

Yes. You can edit an AI Roleplay only in the Draft stage. To edit an AI Roleplay in the Production stage, select the checkbox for the AI Roleplay, and then click Edit in Draft. This creates a draft version while the production version stays unchanged.

You can edit the Scenario, Persona, Evaluation parameters, Exit conditions, and Learner brief.

For more information, see Edit a Roleplay in Mirror.

Unpublish the Workflow that the AI Roleplay is attached to. Unpublishing moves the Workflow and the AI Roleplay to the Draft stage, where you can update the Persona, Exit conditions, or Evaluation parameters of the AI Roleplay. When you publish them again, learners access only the updated version of the AI Roleplay.

Note:

Learners who access a Workflow with Roleplay link that you already shared see the following screen until you publish the Roleplay again:

After you publish it, the same Mirror link remains accessible, so you do not need to share a new link with your learners.

For more information, see Detach a Roleplay in Mirror and Attach a Roleplay in Mirror.

Share a Mirror Roleplay attached to a Workflow as a direct link from the Ready or Production stage of the Whatfix Mirror Dashboard. When you share, select the required Language Variant, and then choose one of the following links:

• Workflow without roleplay: Shares only the Workflow steps, without the Roleplay interaction.

• Workflow with roleplay: Shares the Workflow together with the Roleplay experience.

  

Use the Ready stage to share a Roleplay for internal review, and the Production stage to share it with your learners. Whatfix records Roleplay Analytics only for Roleplays shared from the Production stage.

For more information, see Share a Roleplay in Mirror and Multi-Language Support in Mirror.

Yes, you can export a Mirror Roleplay as a SCORM package to share it through your Learning Management System (LMS).

For more information, see LMS Integration and Export a Mirror Roleplay as a SCORM Package.

No. Mirror Roleplay supports voice and chat interactions only, so a Roleplay session does not require camera access.

Yes. A learner must grant microphone access to begin a voice Roleplay session. Whatfix requests the learner for microphone permission when the session starts.

If the learner denies microphone access, the following message appears. To continue, click Reload to reload the page, and then grant microphone access.

Whatfix supports Mirror Roleplay in Chrome, Firefox, and Microsoft Edge. Whatfix does not support Safari for Mirror Roleplay.

For more information, see Browser Compatibility.

Mirror Roleplay runs directly in the browser, so it works on any Windows or macOS device that runs one of Mirror's supported browsers.

Note:

Mirror Roleplay is not supported on mobile devices or tablets.

For more information, see Browser Compatibility.

Whatfix evaluates each Roleplay attempt against the evaluation parameters you configure and generates an overall score from 0 to 100, along with individual scores for each parameter. Whatfix also generates positive feedback that highlights what the learner did well and constructive feedback that suggests improvements. Review these results on the Roleplay Analytics dashboard.

For more information, see Roleplay Analytics.

Roleplay Analytics does not record an attempt in the following cases:

• The attempt lasts less than one minute.

• You share the Roleplay from the Ready stage. Mirror does not capture or display analytics for Roleplays in the Ready stage.

• Learners have not yet interacted with the Roleplay after you publish and share it from the Production stage.

For more information, see Share a Roleplay in Mirror and Roleplay Analytics.

Yes. There is no limit to the number of times a learner can attempt a Roleplay. Roleplay Analytics records each attempt separately, with its own overall score, evaluation parameter scores, and feedback, so you can compare a learner's performance across attempts.

To view every attempt a learner made, click View details under User level breakdown | Completed Both, and then click the Open icon.

For more information, see Roleplay Analytics.

A learner completes a Roleplay only when they finish both the Workflow and the Roleplay. Roleplay Analytics tracks this with the following events:

• Workflow or Roleplay started: Triggered when a learner starts either the Workflow or the Roleplay.

• Workflow and Roleplay completed: Triggered when a learner finishes both the Workflow and the Roleplay.

Learners who start but do not complete both the Workflow and Roleplay appear in the Attempted list of the User level breakdown, and not in the Completed Both list.

For more information, see Roleplay Analytics and Mirror Events in Trend Insights.

By default, Whatfix identifies learners using browser cookies. If your organization is configured with Einmaliges Anmelden (SSO), Whatfix identifies your learners using the configured User ID.

For more information, see Roleplay Analytics, Set up End-User Authentication and Set up Single Sign On.