Understanding Agent Types in Perspective AI
Perspective AI offers three specialized agent types, each designed for specific conversational scenarios. Whether you're conducting deep research interviews, replacing static forms, or transforming surveys into engaging experiences, choosing the right agent type ensures optimal results.
Agent Types Overview
Interviewer Agent
Purpose: Scales deep, qualitative interviews without losing quality
The Interviewer agent uses ethnographic interviewing techniques to conduct exploratory research. It follows the participant's narrative naturally, letting themes emerge organically rather than following a rigid script. This agent excels at uncovering unexpected insights through adaptive follow-up questions.
Best for:
-
User research and discovery interviews
-
Customer feedback sessions exploring experiences and motivations
-
Exploratory research on new features or concepts
-
Understanding user behaviors and decision-making processes
-
Any scenario where you need rich, narrative data
Concierge Agent
Purpose: Replaces static forms with intelligent conversational flow
The Concierge agent uses an infer-and-validate approach to efficiently collect structured information. It infers details from context, skips questions about information it already knows, and adapts its question order based on previous answers—all while maintaining a natural conversation that never feels like filling out a form.
Best for:
-
Onboarding flows that need contextual information
-
Lead qualification and intake processes
-
Support ticket creation with intelligent triage
-
Any process requiring structured data collection
-
Application or screening processes
Evaluator Agent
Purpose: Turns boring surveys into engaging conversations
The Evaluator agent applies structured assessment methodology to systematically gather feedback on specific criteria. It presents evaluation questions conversationally while maintaining the consistency needed for quantitative analysis, gathering both numerical scores and contextual evidence.
Best for:
-
Product satisfaction surveys (NPS, CSAT, CES)
-
Feature prioritization and structured feedback
-
Usability testing with defined evaluation criteria
-
Post-interaction feedback collection
-
Market research requiring comparable metrics
Choosing the Right Agent Type
Decision Framework
Ask yourself these questions to identify the best agent type:
What's your primary goal?
-
Deep understanding through open exploration → Interviewer
-
Efficient collection of specific information → Concierge
-
Systematic evaluation across defined criteria → Evaluator
What's your question structure?
-
Open-ended, following participant's lead → Interviewer
-
Specific fields with smart inference → Concierge
-
Standardized criteria requiring consistent assessment → Evaluator
What's the participant experience?
-
Exploratory conversation that follows their narrative → Interviewer
-
Quick, guided process that infers what it can → Concierge
-
Structured evaluation that feels conversational → Evaluator
How will you analyze results?
-
Thematic analysis of rich narratives → Interviewer
-
Structured fields for segmentation and analysis → Concierge
-
Quantitative metrics with supporting evidence → Evaluator
Comparison Table
| Feature | Interviewer | Concierge | Evaluator |
|---|---|---|---|
| Primary Use | Exploratory research | Efficient data collection | Structured assessment |
| Question Style | Adaptive, emergent | Inferred and validated | Systematic evaluation |
| Conversation Flow | Follows participant narrative | Progressive narrowing | Criteria coverage |
| Data Output | Rich narratives and themes | Structured, pre-filled fields | Scores with evidence |
| Typical Length | 5-10 questions | 5-10 questions | 5-10 questions |
| Adaptation | Based on insights shared | Based on context and prior answers | Based on scores given |
| Best For | Unknown unknowns | Known information needs | Defined evaluation dimensions |
Common Use Cases by Agent Type
Interviewer Agent Scenarios
New Feature Exploration
Let users share how they'd approach a proposed feature, following their thinking to uncover use cases and concerns you hadn't anticipated.
Customer Journey Understanding
Explore end-to-end experiences by following participants' narratives about their interactions with your product or service.
Churned Customer Research
Understand cancellation reasons through open conversation that follows the participant's story rather than a checklist.
Problem Discovery
Investigate pain points by letting participants guide the conversation to what matters most to them.
Concierge Agent Scenarios
Smart Onboarding
Collect account setup information by inferring details from context (like company domain) and only asking what's truly needed.
Lead Qualification
Gather qualifying information efficiently while pre-filling obvious details and adapting questions based on previous answers.
Support Intake
Route issues intelligently by inferring urgency and details from how participants describe their situation.
Progressive Profiling
Build up user information over time by only asking for new details each interaction, not repeating questions.
Evaluator Agent Scenarios
Post-Purchase Feedback
Systematically assess satisfaction across key dimensions while gathering context about what drove each rating.
Feature Prioritization
Evaluate multiple features consistently, collecting both importance scores and specific evidence about why each matters.
Usability Assessment
Score task completion, ease of use, and satisfaction while capturing concrete examples of what worked or didn't.
Regular Product Health Checks
Track consistent metrics over time while maintaining engagement through conversational presentation.
Creating Your First Agent
Ready to create an agent? Here's what to do next:
-
Identify your use case – Review the scenarios above and determine which agent type fits
-
Read the detailed guide – See the specific article for your chosen agent type:
-
Define your objectives – Clarify what you want to learn or accomplish
-
Create and test – Build your agent and test it yourself before launching
Can You Switch Agent Types?
No—once you select an agent type, it's fixed for that outline. If you realize you need a different agent type, you'll need to create a new outline from scratch.
This is intentional: each agent type uses fundamentally different approaches to conversation. An Interviewer outline optimized for exploratory research won't translate well to Concierge's infer-and-validate methodology, and vice versa.
Best practice: Take time upfront to choose the right agent type based on your goals. Consider:
-
Do I need open exploration? → Interviewer
-
Do I need specific information efficiently? → Concierge
-
Do I need structured evaluation? → Evaluator
If you're unsure, review the use cases and examples for each type before creating your outline.
Best Practices Across All Agent Types
Start with clear objectives. Define what success looks like before creating any agent.
Test yourself first. Experience the conversation as a participant would before inviting others.
Trust the methodology. Each agent type uses proven techniques—let it work as designed.
Keep it focused. All agent types aim for efficiency. Don't overload with unnecessary questions.
Iterate based on real conversations. Review early sessions and refine your approach.