---
title: "Understanding Agent Types"
nav_title: "Agent Types"
description: "What each agent type does, what it replaces, and when to override Perspective's automatic selection."
tags: ["concierge", "interviewer", "evaluator", "advocate", "agent selection"]
date: "2026-03-30"
nav_order: 4
---

# Understanding Agent Types

When you describe what you need, Perspective automatically selects the right agent type for the job. You do not need to choose manually — but understanding what each agent does helps you refine your conversations and know when to override the default.

Each agent type is purpose-built for a different job and uses a distinct methodology. Here is what each one does, what it replaces, and when it is the right fit.

## The Four Agent Types

### Concierge — Replace Your Forms

If you currently use web forms, intake questionnaires, onboarding flows, or any process where you collect structured information field by field, the Concierge agent is for you.

**What it does:** Collects structured data through conversation. Infers details from what people share, skips redundant questions, and adapts in real time — making forms feel effortless instead of tedious.

**Use when you need:** Specific information from people (lead details, support intake, registrations, applications).

[Set up a Concierge agent](/docs/guide/design/concierge-agent)

### Interviewer — Replace Manual Research

If you currently run user interviews, customer calls, focus groups, or any qualitative research where you explore a problem space, the Interviewer agent is for you.

**What it does:** Conducts deep, exploratory research using ethnographic interviewing techniques. Follows the participant's narrative, asks adaptive follow-ups, and lets themes emerge naturally rather than forcing a script.

**Use when you need:** Rich understanding of behaviors, motivations, and unmet needs — especially when you don't know what you don't know.

[Set up an Interviewer agent](/docs/guide/design/interviewer-agent)

### Evaluator — Replace Your Surveys

If you currently send NPS surveys, satisfaction questionnaires, feedback forms, or any tool that collects ratings and scores, the Evaluator agent is for you.

**What it does:** Presents structured evaluation criteria conversationally. Collects quantitative scores with qualitative context, probes after extreme ratings, and keeps participants engaged instead of clicking through a wall of radio buttons.

**Use when you need:** Comparable, quantifiable data across participants — satisfaction scores, feature ratings, prioritization rankings.

[Set up an Evaluator agent](/docs/guide/design/evaluator-agent)

### Advocate — Replace Generic Outreach

If you currently rely on email campaigns, landing pages, sales scripts, or one-size-fits-all messaging to persuade people, the Advocate agent is for you.

**What it does:** Listens first to understand each person's situation, concerns, and priorities, then delivers tailored messaging from your playbook. Handles objections with specific proof points instead of generic pitches.

**Use when you need:** To persuade, educate, or influence — making your case individually to each person.

[Set up an Advocate agent](/docs/guide/design/advocate-agent)

## Decision Matrix

| | **Concierge** | **Interviewer** | **Evaluator** | **Advocate** |
|---|---|---|---|---|
| **Goal** | Collect information | Discover insights | Measure and score | Persuade and influence |
| **Replaces** | Forms and intake flows | Manual interviews and research calls | Surveys and feedback forms | Generic outreach and pitches |
| **Output** | Structured data (fields, values) | Qualitative themes and narratives | Quantitative scores with context | Personalized conversations that move people |
| **Question style** | Infer and validate | Open-ended and adaptive | Structured criteria with conversational tone | Discovery first, then tailored advocacy |
| **Best for** | Onboarding, lead qualification, support intake, registrations | User research, customer discovery, exit interviews, concept exploration | NPS, CSAT, feature prioritization, usability evaluation | Product advocacy, sales objection handling, policy communication, cause-driven outreach |
| **Conversation length** | 5-10 questions | ~10 questions | 5-10 questions | Varies by engagement |

## How to Decide

**Start with your goal, not your current tool.**

- "I need to collect specific data from people" — **Concierge**
- "I need to understand something I don't fully grasp yet" — **Interviewer**
- "I need to measure something consistently across many people" — **Evaluator**
- "I need to convince people of something" — **Advocate**

**If you are unsure between two:**

- Concierge vs. Evaluator: Do you need freeform data (Concierge) or ratings and scores (Evaluator)?
- Interviewer vs. Evaluator: Are you exploring open questions (Interviewer) or measuring known criteria (Evaluator)?
- Interviewer vs. Advocate: Are you trying to learn from people (Interviewer) or influence them (Advocate)?
- Advocate vs. Concierge: Are you trying to persuade (Advocate) or collect information (Concierge)?

## Agent Type is Set at Creation

The agent type is locked once your conversation is created — each type uses a fundamentally different methodology, so switching after the fact is not supported. If you need a different type, create a new conversation.

In practice this rarely matters, because Perspective's automatic selection is usually correct. If you are unsure, just describe what you need and let Perspective decide.