The Research Agent allows you to enrich user or account profiles by sending requests in natural, human-like language - as if you were talking to a person. Based on your prompt, the agent returns structured data that can be used to update existing attributes or create new ones for a selected user or account.
Creating a Research agent
Go to the Agents section and select "Create agent" -> "Research agent".
Next, choose the data type the agent will work with: User or Account.
In this guide, we’ll create a Research Agent for users, but the process is the same for accounts.
Once the agent is created, you’ll see the Research Agent editor. Below is a breakdown of the main sections:
Question section. A single prompt the agent will process.
Prompt. The content of your question. It should be written in clear, natural language.
Output type. Choose an existing attribute or create a new one to store the agent's response.
Attribute. The specific attribute that will be updated after the agent runs.
Add a question. You can define multiple questions per agent. All questions will be executed in a single run.
Test. Select a sample user or account to test the agent and ensure your prompt works as intended.
Apply Filters (Optional)
You can define which users or accounts the Research Agent should run for by applying filters. These filters can contain just attributes, so you can't apply filters based on user/account behaviour.
📘 Good to know
If a user doesn't match the filters defined in the Research Agent but reaches a journey step where the agent is used, the journey flow will end for that user at that point.
Once your agent is created and tested, you can start using it in your journeys to dynamically enrich user or account data!