Intro to Roleplay Persona Customization
Persona customization lets you control how a roleplay agent behaves through structured settings, instead of relying only on the prompt. You'll find it in the Behavior block of the Persona section in the agent editor (Practice > Roleplay Agents, open an agent, then click Edit Agent). You can tune:
- Information strategy: how much the agent shares, when, and how directly.
- Engagement progression: how the conversation moves from rapport to credibility to pain acknowledgment.
- Difficulty: how challenging the agent is, including objections, skepticism, patience, and complexity.
- Adaptive difficulty: whether difficulty adjusts during the call based on how the team member is doing.
- Conversation style: verbosity, response length, storytelling, humor, and how often the agent asks questions.
Why use it?
- Consistent challenge: make agents realistically challenging so team members get predictable practice.
- Standardized difficulty: align practice difficulty across teams and regions.
- Buyer-type tuning: match agent behavior to different buyer types, such as a skeptical executive, a curious evaluator, or a price-focused prospect.
- Beyond prompts: control behavior through structured settings instead of relying only on free-form prompt text.
Important product details
- Behavior settings apply to the main roleplay agent (the buyer or decision-maker) and its scorecard. If the agent has a gatekeeper, the gatekeeper phase uses its own persona and a separate scorecard, and behavior settings don't change it.
- The Background Environment setting applies to the gatekeeper when the agent has one. If there is no gatekeeper, it applies to the main agent. See Create a Gatekeeper to a Roleplay for details.
Choosing or creating a behavior
The Behavior block has two options:
- Existing Behavior: pick a saved behavior from the list of behavior cards, for example Conversationalist, Analyst or Skeptic.
- Custom Behavior: start from the current settings, give the behavior a name, and adjust the settings. A custom behavior shows a Saved with this agent badge.
With Custom Behavior selected, click Edit Settings to open the Behavior window. Give the behavior a name in Behavior Name, adjust the settings, and click Save Behavior. Click Reset Changes (or Reset to the original behavior's name) to undo your edits, or Cancel to close the window without saving. Next to a changed setting, Reset puts that one setting back to its original value.
The settings are organized into five groups: Information Strategy, Engagement Progression, Difficulty, Adaptive Difficulty, and Conversation Style. Back in the editor, each group shows a short summary, and a changed badge shows how many settings in that group you changed. Most sliders run from 1 (low) to 10 (high).
Information Strategy
Controls how the agent shares information and when it surfaces pain points.
Setting | What it does |
|---|---|
Proactive Sharing | How much the agent volunteers without being asked: Minimal, Moderate or Open. |
Direct Answer | How directly the agent answers questions. Higher values mean fewer hedges and redirects. |
Deflection Tendency | How often the agent sidesteps questions. High values simulate evasive contacts. |
Detail Level | How much detail the agent gives: High Level, Detailed or Granular. |
Pain Point Timing | When the agent reveals pain points: Immediate, After Rapport or After Trust. |
Pain Point Depth | How openly the agent talks about its pain: Resistant, Moderate or Open. |
Example: for a skeptical executive, set Deflection Tendency higher and Proactive Sharing to Minimal so the team member has to earn information. For a curious evaluator, set Proactive Sharing to Open and Detail Level to Detailed or Granular.
Engagement Progression
Controls how the conversation moves through stages of engagement.
Setting | What it does |
|---|---|
Initial State | How the agent starts the conversation: Closed, Neutral or Receptive. |
Verbal Demonstration, Rapport Building, Credibility Establishment, Pain Acknowledgment, Personalization | Each has one slider (Low to High) for how much that factor opens the agent up. Higher means a bigger effect. |
Progression Stage | The engagement stage the agent aims for: Guarded, Curious, Engaged or Collaborative. |
Regression Triggers | What makes the agent step back: Pushiness, Irrelevance or Time Wasting. |
Example: for price-focused prospect training, set Credibility Establishment high so the agent only acknowledges pain once value is clear, and set Regression Triggers to Pushiness so a premature discount pitch pushes the agent back.
Difficulty
Controls how challenging the agent is.
Setting | What it does |
|---|---|
Overall Difficulty Level | The overall challenge of the session. Higher levels combine tougher objections, more withholding and stronger skepticism. |
Objection Sophistication | Low means simple, single objections. High means compound objections with several underlying concerns. |
Information Withholding | How much relevant context the agent keeps back until the team member asks the right questions. |
Skepticism Level | How skeptical the agent is. |
Patience Threshold | How long before the agent shows impatience or disengages. |
Decision Complexity | Low means the agent can decide alone. High means multiple stakeholders, approvals or budget cycles. |
Example: for consistently challenging practice, set Overall Difficulty Level and Objection Sophistication high.
Adaptive Difficulty
Controls whether difficulty adjusts during the call.
Setting | What it does |
|---|---|
Adaptive Difficulty | On or Off: when on, the agent gets easier or harder based on how the team member is doing. |
Adjustment Speed | How quickly difficulty changes: Slow, Moderate or Fast. |
Performance Window | How much recent performance (Short to Long) is used when adjusting difficulty. |
Example: for self-serve practice where one agent serves several skill levels, turn Adaptive Difficulty on.
Conversation Style
Controls how the agent speaks.
Setting | What it does |
|---|---|
Verbosity | How much the agent says per turn: Brief, Moderate or Detailed. |
Response Length Preference | Brief, Balanced or Detailed responses. |
Storytelling Tendency | How often the agent uses stories, examples or analogies. Higher values give the team member more to listen to. |
Humor Usage | How much humor the agent uses. |
Question Asking Frequency | How often the agent asks questions back: Minimal, Moderate or Inquisitive. |
Example: for executive-style practice, set Verbosity and Response Length Preference to Brief. For a collaborative evaluator, set Question Asking Frequency to Inquisitive and raise Storytelling Tendency.
Examples
Example 1: Skeptical executive (hard)
- Information Strategy: Proactive Sharing Minimal, Direct Answer low, Deflection Tendency high, Detail Level High Level, Pain Point Timing After Trust, Pain Point Depth Resistant.
- Engagement Progression: Initial State Closed, high Credibility Establishment, Regression Triggers Pushiness.
- Difficulty: Overall Difficulty Level high, Objection Sophistication high, Skepticism Level high, Decision Complexity high.
- Conversation Style: Verbosity Brief, Response Length Preference Brief, Humor Usage low, Question Asking Frequency Inquisitive.
Example 2: Curious evaluator (medium)
- Information Strategy: Proactive Sharing Open, Direct Answer high, Deflection Tendency low, Detail Level Detailed, Pain Point Timing After Rapport, Pain Point Depth Open.
- Engagement Progression: Initial State Receptive, high Rapport Building and Verbal Demonstration.
- Difficulty: Overall Difficulty Level medium, Skepticism Level low to medium, Patience Threshold high.
- Conversation Style: Verbosity Moderate, Storytelling Tendency high, Question Asking Frequency Inquisitive.
Example 3: Price-focused prospect (medium to hard)
- Information Strategy: Proactive Sharing Minimal, Deflection Tendency high, Pain Point Timing After Rapport, Pain Point Depth Moderate.
- Engagement Progression: high Credibility Establishment, Regression Triggers Pushiness.
- Difficulty: Information Withholding high, Objection Sophistication medium, Patience Threshold medium.
- Conversation Style: Response Length Preference Balanced, Question Asking Frequency Moderate.
Example 4: Adaptive difficulty (self-serve)
- Difficulty: Overall Difficulty Level medium as the starting point.
- Adaptive Difficulty: Adaptive Difficulty on, Adjustment Speed Moderate.
- Other settings for a typical discovery call. Difficulty then moves up or down based on how the team member performs during the session.
Tips
- Click Save Behavior, then save the agent (Save Draft or Publish) after changing behavior settings. Changes apply to new roleplay sessions.
- Use Instructions for Team Members or the scenario text to set the team member's expectations so they match the behavior you configured.
- To reuse a behavior on another agent, choose Existing Behavior and pick it from the list. You can also clone an agent that already has the behavior you want and change its name, scenario and scorecard.
For gatekeeper setup, scorecards and background noise, see Create a Gatekeeper to a Roleplay. For multi-persona roleplays, see Create a Multi-Persona Roleplay Agent.
Hope this article was helpful. For any queries, please write to support@outdoo.ai