The Glean Company Search Action searches across company data. This action calls the Glean search engine and can issue any query that you can issue against the Glean search engine, helping you query internal company knowledge. The Glean Company Search Action:
  • Queries Across Company Data: Provides a targeted search across various data sources within the company, including internal documents, databases, and content repositories.
  • Adds Search Results to Agent’s Memory: The results from these searches are loaded into the agent’s memory for use in subsequent steps or analyses. This action does not output anything to the user—you will need to follow it with a Respond step with instructions on how to summarize or extract information from the search results and format them for the user.
  • Supports Multiple Knowledge Sources: Enables searches over multiple designated knowledge sources or repositories, such as specific folders or databases.

​
Usage Examples

  • Meeting Preparation: Pulls the latest company notes, updates from internal sources like Salesforce, and meeting transcripts to prepare for meetings.
  • Candidate Interview Preparation: Searches a candidate’s submitted resume and internal interview notes to curate relevant information for interviews.
  • Support Ticket Resolution: Searches Confluence pages for playbooks relevant to solving a support ticket.

​
Configuration

You can configure the Glean Company Search Action using the Agent Builder tool with the following steps:
  1. Define Search Queries: You can specify search queries manually or provide natural language instructions for the AI to determine appropriate queries.
    • Manual Specification: You can list a template for any search query that you can enter into the Glean search engine. These templates can also include input fields. Examples:
      1. app:gmail from:me to:”Arvind Jain”
      2. [[Feature name]] app:jira updated:past_week
      3. [[Customer name]] app:salescloud type:opportunity isclosed:false
    • Natural Language Instructions: You can provide instructions to the AI to figure out the necessary search queries. Examples:
      • Find all Slack messages that mention [[Project name]]
      • Search Confluence for playbooks that are relevant to solving the issue in [[Zendesk ticket]]
  2. Filter Sources: Optionally specify that the search should be issued against a specific set of data sources, containers (e.g., folders, collections), or documents.
  3. Select Number of Results to Return: Choose how many results to return per search. Fewer results will result in a more focused response with less breadth.
  4. Select How Much Information to Return: Choose to either return only relevant snippets in results or return the entire document containing the result.
The configured search results can be dynamically integrated into agent workflows to provide comprehensive data for analysis and decision-making.