Skip to main content

Expert Workers

Expert workers (also called expert agents) are specialized agents that you can attach to your main agent to handle specific types of files or tasks. They allow you to delegate complex processing to purpose-built AI assistants with their own prompts and model configurations.

What are Expert Workers?​

Expert workers are organization-level agents that focus on handling specific file types or specialized tasks. Rather than having a single agent handle all file types with generic instructions, you can create expert workers with specialized knowledge and instructions for particular domains—like image analysis, PDF processing, code review, or document summarization.

When a file is uploaded to your main agent, the system can route it to the appropriate expert worker based on the file type. This allows you to leverage different AI models, custom prompts, and specialized tools for different kinds of content.

When to Use Expert Workers​

Expert workers are useful when you need:

Specialized handling for specific file types. For example, you might have one expert worker for analyzing images, another for processing PDFs, and a third for reviewing code files.

Different AI models for different tasks. You might use a vision-capable model for images but a text-focused model for documents, all within the same agent.

Custom instructions tailored to specific file categories. Each expert worker can have its own system prompt with domain-specific guidelines and best practices.

Domain expertise in focused areas. Expert workers let you leverage the best AI capabilities for each type of content users upload.

Creating an Expert Worker​

To create a new expert worker, navigate to your organization settings and go to the Expert Workers section. Click Create Expert Worker and configure:

Title - A descriptive name for the expert worker (e.g., "Image Analyst", "PDF Processor")

Description - A brief explanation of what this expert worker handles

Prompt - Custom instructions for how this expert worker should process files. This is where you provide specialized guidance specific to the file type.

AI Provider - The service providing the AI (e.g., OpenAI, Google, Anthropic)

Model - The specific AI model to use for this expert worker

Supported File Types - Select the file categories this expert worker should handle:

  • Images (JPG, PNG, etc.)
  • PDFs
  • Text files (TXT, MD, etc.)
  • Audio files
  • Video files
  • Word documents
  • Excel spreadsheets
  • PowerPoint presentations
  • CSV files
  • Calendar files

You can create multiple expert workers handling the same file type if you need different approaches for different scenarios.

Attaching Expert Workers to Agents​

Once you've created expert workers, you can attach them to your agents in the template editor. In the Expert Workers sidebar tab, click Add Expert Worker to select which expert workers should be available for that agent.

When a user uploads a file to your agent:

  1. The system identifies the file type
  2. It checks which expert workers handle that file type
  3. The file is routed to an appropriate expert worker
  4. The expert worker processes the file with its specialized prompt and model
  5. The results are returned to the main conversation

Example: Multi-Expert Document Analysis​

Imagine you have an agent for document analysis. You could create three expert workers:

  • Image Expert - Handles PNG, JPG files with vision-focused instructions for analyzing diagrams, screenshots, and visual content
  • PDF Expert - Processes PDF documents with instructions for extracting structured information and handling complex layouts
  • Text Expert - Reviews plain text and markdown files with instructions for analyzing code, documentation, and written content

When a user uploads a file, the right expert automatically handles it based on its type, providing specialized analysis tailored to that content format.

Best Practices​

Keep prompts focused - Each expert worker should have clear, specialized instructions for its file type. Avoid trying to make one expert worker handle too many different scenarios.

Use appropriate models - Select AI models suited to the task. Use vision-capable models for image processing, and text-focused models for document analysis.

Test with sample files - Before deploying, test each expert worker with representative files to ensure it provides the quality of analysis you need.

Document expertise - In the expert worker's description and prompt, explain what it specializes in so other team members understand when to use it.

Avoid overlapping handlers - If multiple expert workers handle the same file type, make sure their prompts are distinct so users understand which one they're using.