---
title: "Your First Conversation in 5 Minutes"
nav_title: "Quickstart"
description: "Create, test, share, and analyze your first Perspective conversation."
tags: ["quickstart", "getting started", "first conversation"]
date: "2026-06-02"
nav_order: 0.5
nav_display: true
---

# Your First Conversation in 5 Minutes

This quickstart walks through the core Perspective loop: create a conversation, test it yourself, invite participants, and analyze the first responses.

## 1. Create the conversation

Open Perspective, click **New agent**, and describe the outcome, audience, and context in plain language:

```text
I want to interview recent customers about their onboarding experience
so I can identify where they got stuck and what helped them succeed.
```

Perspective selects an agent type, drafts the outline, and sets up the flow. You can accept the draft or ask the design agent to refine it.

For more detail, see [Create a Conversation](/docs/guide/design/create-conversation) and [Conversation Outline](/docs/guide/design/create-research-outlines).

## 2. Check the agent type

Perspective chooses the agent from your description:

| Agent           | Best for                                                                         |
| --------------- | -------------------------------------------------------------------------------- |
| **Concierge**   | Intake, qualification, lead capture, support triage, and form replacement        |
| **Interviewer** | Exploratory research, discovery, user interviews, and feedback conversations     |
| **Evaluator**   | Surveys, scoring, rubrics, and structured feedback with follow-up questions      |
| **Advocate**    | Personalized outreach, objection handling, education, and campaign conversations |

You can override the agent type if the automatic choice does not match your goal. See [Understanding Agent Types](/docs/guide/design/choosing-agents).

## 3. Test it yourself

Before sending the conversation to real participants, open the design view and select **Try it out**.

Complete the preview like a participant would. Watch for:

- Questions that feel unclear or repetitive
- Missing context the agent should know
- A tone that feels too formal or too casual
- A conversation that runs too long
- An ending that does not explain what happens next

If something feels off, tell the design agent what to change in plain language, then test again. See [Test and Refine](/docs/guide/design/test-and-refine).

## 4. Invite participants

When the preview feels ready, open **Invite participants**.

You can:

- Copy the public conversation link
- Send targeted invitations by email
- Invite through Slack, WhatsApp, or phone when those channels are configured
- Embed the conversation on a website or app

For all collection options, see [Invite Participants](/docs/guide/collect/invite-participants), [Embed Interviews](/docs/guide/collect/embed-interviews), and [Track Participant Context](/docs/guide/collect/track-participant-context).

## 5. Analyze the first responses

After at least one non-preview conversation is complete, open the Perspective's Results area and select **Analysis**.

Ask a plain-language question:

- "What were the biggest onboarding pain points?"
- "Summarize the strongest themes with citations."
- "Which participants mentioned pricing or setup confusion?"
- "Create a table of the top feature requests."
- "Show a heatmap of pain points by segment and journey stage."

Analysis sessions can search transcripts, use form data, create charts or tables, cite source conversations, and save useful outputs as highlights or public pages. See [Analysis Sessions](/docs/guide/analyze/analysis-sessions) and [Conversation Results](/docs/guide/analyze/conversations).

## What to read next

- [Design Conversations](/docs/guide/design) for outline, agent, setting, and completion-flow setup
- [Collect Responses](/docs/guide/collect) for links, embeds, participant groups, and context
- [Analyze and Share](/docs/guide/analyze) for results, trust assessment, analysis sessions, and pages
- [Developers](/docs/build) for MCP, API, webhooks, automations, and embed reference