Persona Evaluation
Persona Evaluation is an automated quality check that reviews your configuration — Business Context, Tone of Voice, and Guidances. It ensures that the Persona guiding your AI Agent is clear, consistent, and actionable, so the Agent can respond to customers accurately, confidently, and within its designed capabilities.
The evaluation helps identify:
- Missing or unclear goals
- Confusing or conflicting instructions
- Overly long or non-actionable descriptions
- Tone of voice inconsistencies
- Instructions that might cause the Agent to hallucinate or loop
The output includes a rating, comments, and suggested improvements for each evaluated section.
What are the Key Evaluation Criteria
The Persona Evaluation checks your configuration against several core dimensions of quality and clarity:
| Criteria | What It Cheks |
|---|---|
| Clarity of Persona Identity | Is the Agent’s role and purpose within your company clearly defined? |
| Clarity of Persona Goal | Are the Agent’s objectives specific, operational, and achievable with its tools? |
| Clarity of Persona Tone of Voice | Is the intended communication style clearly described and actionable? |
| Conciseness of Instructions | Are the Persona and Guidances concise and free of redundant or verbose wording? |
| Clarity of Execution Instructions | Are all instructions actionable, operational, and easy for the AI to follow? |
| Redundancy and Actionability | Is there any poetic, vague, or repetitive language that could confuse the AI? |
| Conflicting Instructions | Are there contradictions (e.g., “never transfer to an agent”) or instructions that conflict with Reasoning default tools (e.g. transfer_to_agent tool)? |
| Hallucination-Prone Instructions | Are there any instructions that could make the AI generate information without a valid source? |
| Implicit Knowledge Assumption | Does the Persona assume access to knowledge or context that the AI doesn’t actually have? |
| Persona Identity Confusion | Does the Persona incorrectly refer to “customer support” or “the company” as external entities instead of speaking as part of them? |
Each area is scored as Good, To Improve, or Poor (displayed in the UI as color-coded levels)
How to Use in Practice
Follow these simple steps to make the most of Persona Evaluation:
Step 1: Add Your Configuration
Define your Business Context and Tone of Voice according to your business needs.
Make sure you include:
- Your company name and what your AI Agent does
- The Agent’s role (e.g., “customer support assistant”)
- A clear tone that reflects your brand (e.g., professional, empathetic, concise)
Step 2: Run the Evaluation
Click “Run Evaluation” in the top-right corner of your Persona configuration screen.
You can also trigger the same check automatically when you save your Persona or Guidance.
Step 3: Review the Results
The evaluation will show:
- A rating for each quality area (from Poor to Good)
- Comments explaining the reasoning
- Actionable suggestions and ready-to-use rewritten examples of your Persona and Guidances
Step 4: Adjust and Apply
Review the proposed changes carefully — they are recommendations, not automatic fixes.
Apply your own common sense and business judgment before accepting any edits.
For example, you may choose to adjust phrasing or keep certain brand-specific elements.