Skip to main content
Personas are AI-generated individuals who participate in conversations with your agents during simulations. Each persona has unique attributes, personality traits, and behavioral patterns that create realistic and diverse interactions.

Overview

Personas represent the people your agent will interact with in real-world scenarios. They’re automatically generated based on your simulation’s scenario and provide the diversity needed for comprehensive agent evaluation.

How Personas Work

Each persona is automatically generated with a unique profile that includes:
  • Summary - A brief descriptive label like “Detail-Oriented Analyst” or “Frustrated Long-time Customer”
  • Story - Background context that explains their situation and needs
  • Purpose - Their specific goal within the simulation scenario
  • Attributes - Detailed personality traits, communication style, and behavioral patterns
Personas go through an approval process where you can review and approve them before they participate in conversations. This ensures only realistic, appropriate personas interact with your agent.

Approval Workflow

1

Generation

Personas are automatically generated based on the simulation scenario
2

Review

If auto_approve is false, personas require manual review and approval
3

Approval

Approved personas can participate in conversations
4

Assignment

Approved personas are assigned to conversations with the agent

Approval Status

  • pending: Persona has been generated and awaits review
  • approved: Persona has been approved and can participate in conversations
  • rejected: Persona has been rejected and will not participate

Example Personas

Customer Support Scenario

Persona Generation

Personas are generated automatically based on:
  1. Simulation Scenario - Provides context for the types of people who would be involved
  2. Target Diversity - Ensures varied personality types, technical levels, and communication styles
  3. Realistic Distribution - Balances common persona types with edge cases

Persona Attributes

Persona attributes are structured into several categories:
  • Age Group: teen, young_adult, adult, or senior
  • Education: Educational background and qualifications
  • Occupation: Current job or profession
  • Economic Status: Financial situation and spending patterns
  • Personality Traits: Core characteristics and behavioral tendencies
  • Values: Fundamental principles they live by
  • Habits: Specific behavioral patterns that reveal character
  • Interests: Main passions or areas of expertise
  • Speech Style: Overall approach to conversation and interaction
  • Speech Patterns: Specific patterns with examples of typical phrases
  • Typical Behavior: How they act in social situations, work, and daily life
  • Stress Triggers: Situations that create tension for them
  • Stress Reactions: How they cope with or react to stress

Best Practices

Review generated personas before approval to ensure they align with your testing goals and represent realistic user types.
Persona diversity is automatically managed, but you can reject personas that don’t fit your evaluation needs.
Approved personas directly impact conversation quality - reject personas that seem unrealistic or inappropriate for your scenario.
The approval process allows you to fine-tune the types of interactions your agent will face during evaluation.