Skip to main content
A conversation is an interview between your Researcher Agent and a user. The agent asks questions based on your study goals and adapts follow-ups based on user responses.

​
How conversations work

  1. User opens link: They see your outreach message and can start chatting
  2. Agent asks questions: Based on your study goals, the agent guides the conversation
  3. Adaptive follow-ups: The agent goes deeper on interesting responses
  4. Natural conclusion: The conversation ends when goals are covered or the user stops responding
Each conversation is unique. Your Researcher Agent adapts its questions based on what the user shares, ensuring natural and relevant discussions.

​
Conversation flow

Every conversation follows this structure:
  • Outreach message: Your custom first message that greets the user
  • Goal-driven questions: Questions designed to address your study objectives
  • Follow-up questions: Deeper exploration based on user responses
  • Closing: A natural wrap-up when the conversation is complete

​
Conversation summaries

Each conversation gets an AI-generated summary displayed at the top of the transcript view. Summaries let you quickly assess what was discussed without reading the full transcript. Hover over View full summary to see the key bullet points.

​
PII detection

Personally identifiable information is automatically redacted in conversations. Redacted content appears as inline badges indicating the type of data removed (names, emails, phone numbers, addresses, credit cards, financial info, SSNs, ID documents, and dates of birth). This keeps transcripts readable while protecting your users’ privacy.

​
Voice messages

Participants can send voice messages instead of typing. Voice messages are automatically transcribed so they appear as text in the transcript alongside the original audio. Click the play button on any voice message to listen to the original recording — you see playback controls with a scrub bar and time display.

​
What makes a great conversation

Not all conversations are equal. The most valuable interviews share these traits:
  • User shares detailed, thoughtful responses
  • Multiple study goals are addressed
  • Conversation reveals unexpected insights
  • User provides specific examples or stories
Use conversation summaries to quickly spot high-value conversations without reading every transcript.

​
Conversation metadata

Each conversation captures contextual information:
  • Custom metadata: Any data you pass via URL parameters
  • Dynamic persona: Per-conversation persona parameters captured at launch
  • Location: Country, region, city (from IP)
  • Device: Desktop, mobile, tablet
  • User-agent: Browser information
  • Locale: User’s language setting
This metadata helps you segment and filter conversations during analysis.

​
Next steps

Customize outreach

Write effective opening messages.

Customize metadata

Track user attributes via URL parameters.

Review conversations

Learn how to navigate and read conversations.

Insights

See how conversations become insights.