Overview

A persona is a simulated customer character used to test your voice AI agents. Each persona represents a different type of caller with specific characteristics - from their language and accent to their emotional state and communication style.
Navigate to Simulations → Personas in your Roark dashboard to view and manage your personas.
Personas List

Creating a Persona

To create a new persona, navigate to Simulations → Personas in your dashboard and click the “Create New Persona” button. The creation modal is organized into collapsible sections where you’ll define who your simulated customer is and how they behave during calls.

Basic Information

Define the fundamental characteristics of your persona - who they are and their calling environment.
SettingOptionsDefault
NameAny descriptive textRequired
LanguageEnglish, Spanish, German, Hindi, French, Dutch, Arabic, GreekEnglish
AccentUS, US Southern, British, Australian, Spanish, German, Indian, French, Dutch, Saudi Arabian, GreekUS
GenderMale, Female, NeutralFemale
Background NoiseNone, OfficeNone

Speech Settings

Control how your persona speaks - their pace, clarity, and natural speech patterns.
SettingOptionsDefaultImpact
Speech PaceSlow, Normal, FastNormalHow quickly the persona speaks
Speech ClarityClear, Vague, RamblingClearHow directly they express themselves
DisfluenciesEnabled, DisabledDisabledAdds “um”, “uh” and natural hesitations

Behavior Settings

Define the persona’s emotional state and communication style during the conversation.
SettingOptionsDefaultImpact
Base EmotionNeutral, Cheerful, Confused, Frustrated, Skeptical, RushedNeutralUnderlying emotional tone
Intent ClarityClear, Indirect, VagueClearHow directly they state their goal
Confirmation StyleExplicit, VagueExplicitHow clearly they confirm things
Memory ReliabilityHigh, LowHighConsistency of information provided

Backstory

Optional text field for additional context about the persona’s background and situation. Helps create more nuanced interactions.

Example Personas

Below are sample personas that demonstrate common customer types you might encounter. Each example shows the configuration settings and explains when to use that persona type for testing.

The Frustrated Regular

Name: "Sarah - Billing Issue"
Language: English
Accent: US
Gender: Female
Background: Office
Speech Pace: Fast
Speech Clarity: Clear
Base Emotion: Frustrated
Intent Clarity: Clear
Backstory: "Long-time customer calling about third billing error this year. 
           Usually patient but reaching limit."
Use for: Testing de-escalation, problem resolution, empathy responses

The Confused Elder

Name: "Robert - Tech Support"
Language: English
Accent: US
Gender: Male
Background: None
Speech Pace: Slow
Speech Clarity: Vague
Disfluencies: Enabled
Base Emotion: Confused
Intent Clarity: Vague
Memory Reliability: Low
Backstory: "Retired, not tech-savvy, calling about password reset. 
           May repeat questions or forget earlier answers."
Use for: Testing patience, clear explanations, handling repetition

The Happy Path Customer

Name: "Alex - Quick Inquiry"
Language: English
Accent: British
Gender: Neutral
Speech Pace: Normal
Speech Clarity: Clear
Base Emotion: Cheerful
Intent Clarity: Clear
Backstory: "Calling to check order status. Has order number ready."
Use for: Baseline testing, happy path scenarios, quick resolutions

The Busy Professional

Name: "Maria - Appointment Change"
Language: Spanish
Accent: Spanish
Gender: Female
Background: Office
Speech Pace: Fast
Speech Clarity: Clear
Base Emotion: Rushed
Confirmation Style: Vague
Backstory: "Calling during work break, needs to reschedule quickly. 
           May not fully listen to options."
Use for: Testing efficiency, handling rushed callers, clear communication

Using Personas

Once created, personas can be:
  • Selected when creating new simulations
  • Reused across multiple test scenarios
  • Modified to create variations for different tests
  • Shared with your team for consistent testing
Start with 3-5 core personas representing your most common customer types, then expand based on edge cases you discover.

Reference

Next Steps