Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.usepropane.ai/llms.txt

Use this file to discover all available pages before exploring further.

Research agents are available on the Growth plan. See Set up your workspace for plan details.
Propane allows you to create conversational research agents that adapt to user responses in real-time. The AI agent drives dynamic conversations to gather deeper insights from your users.

Step-by-Step Guide

1

Initialize the research agent

  1. Navigate to Agents from the sidebar navigation.
  2. Click the New button in the top right corner.
  3. Select your template type from the dropdown menu (e.g., NPS Follow up, Win/Loss, etc.).
2

Configure research agent details

Once on the Create page, fill in the foundational details:
  • Name: Give your research agent a distinct internal name (e.g., “Q3 NPS Follow-up” or “New User Sentiment”).
  • Language: Select the language in which the session will be conducted.
  • From: Choose the sender profile. This determines the name and email address participants will see when they receive the invitation.
3

Set agent settings

This section controls the personality and focus of the AI agent.
  • Opening Message: This is the specific hook the agent uses to start the chat. Propane provides a default, but you can customize it using variables like {{participant_name}}.
  • Advanced Settings > Additional Instructions: This is where you define the “brain” of the research agent. You can tell the AI how to behave, what to prioritize, and what to avoid.

Leveraging NPS Scores in Instructions

Because the AI agent knows the participant’s score (0-10) before the chat begins, you can write conditional instructions to tailor the experience.Examples of Effective Instructions:
GoalExample Instruction
Tailor by Score”If the user is a Detractor (score 0-6), apologize for the experience and ask what the one thing we could fix is. If they are a Promoter (9-10), ask them to describe their favorite workflow.”
Dig Deeper on Features”If the user mentions ‘UI’ or ‘Design’, ask specific follow-up questions about which buttons or menus were confusing. Do not settle for general answers.”
Manage Tone”Adopt a casual, empathetic tone. Keep the conversation short (max 3-4 questions).”
Blank Slate Approach”Treat this session as a fresh start. Do not assume knowledge of previous conversations outside of the current NPS score.”
4

Customize landing page

Configure the text that participants see before and after the chat:
  • Welcome text: The message displayed on the landing page before the session begins.
  • Thank you text: The message displayed once the participant completes the session.
You can use dynamic variables like {{participant_name}} in these fields to personalize the landing page experience.
5

Launch

  1. Review your settings to ensure the instructions align with your goals.
  2. (Optional) If you wish to offer rewards for completion, configure the Incentives section.
  3. Click the Create button in the bottom right corner.
Once created, you will be redirected to the Agents dashboard where a success message will appear. Your new research agent will be listed in the grid, ready for you to add participants.