Autogen Agents Taking Over? Trying To Create a Ridiculous Amount of Agents - Real Business Use Case



AI Summary

  • Introduction to building a “unicorn” with autogen
    • Inspired by a video on autogen Auto Builder
    • Idea to create multiple agents and store them in a JSON file
    • Use autogen to select relevant agents for tasks
  • Yesterday’s progress
    • Created agents: Director, Consumer CMO, Cold Email Specialist, Proofreader
    • Recommended watching the previous video for context
  • Today’s goal: Build an enterprise with various departments
    • Created SEO team with descriptions for Junior, Senior, Critic, and Director roles
    • Converted descriptions into a JSON file
  • Expansion to other departments
    • Marketing, Legal, Product, Finance, Data Science, HR, Sales, CSIT, R&D, Operations, PR, Strategy, BSEV Compliance, QA, Sustainability, and CSR
    • Aimed to create a Junior, Senior, Critic, and Director for each department (44 agents total)
  • Encountered an error due to too many roles
    • Reduced the number of departments and roles per department
    • Focused on Senior, Critic, and Director roles
  • Task execution with selected agents
    • Built a comprehensive business plan for “add scale”
    • Included legal, marketing, finance, and sales planning
  • Agent selection process
    • Initially selected C-level agents
    • Adjusted to include seniors and critics from each department
  • Understanding the agent selection logic
    • Explored the agent Builder python file
    • Logic based on task complexity and agent roles
  • Attempted to include user input in agent selection
    • Tried to add a “user agent” to the list
    • Plan to research and understand the agent Builder file further
  • Final thoughts
    • Interested in building a large pool of agents for tasks
    • Seeking feedback on incorporating human input into the script
    • Encourages likes, comments, and subscriptions for content direction