Using Bot Framework + AutoGen together : What is your vision ?

[QaisAlkhateeb]

Hello everyone,

Over the past few years, my team and I have built several conversational AI bots using the Microsoft Bot Framework. We typically rely on it for conversation orchestration, dialog management, and multi-channel deployment. Most of our AI-driven features were powered by Bot Framework’s standard LUIS/QnA integration or custom ML models, and it worked well for structured tasks and straightforward interactions.

Now, we’re exploring the possibility of combining Bot Framework with AutoGen to handle more sophisticated generative reasoning and autonomous decision-making. Essentially, we’d like the Bot Framework to keep doing what it does best—managing the user interaction flows—while AutoGen handles reasoning part and generate responses (using GenAI and Tools integration).

Before fully committing to this path, I’d love to validate my understanding:

Primary Responsibilities

• Bot Framework: Conversation flow, dialog management, and channel integration.

• AutoGen: Advanced generative AI, chain-of-thought, or task-specific reasoning.

Proposed Integration

• The Bot Framework would handle all incoming messages and orchestration.

• When complex problem-solving or creative text generation is needed, it would hand off to an AutoGen-based system, then return the results back to the user via the Bot Framework.

Questions

I’d love to validate this approach:

• Do you think this separation of concerns (conversation vs. advanced generative reasoning) is the right approach?

• Could the Bot Framework become obsolete with the new features and updates of AutoGen?

• Is there any plan to replace the Bot Framework entirely with AutoGen?

• What is your overall vision for the future of these two solutions?

• If you were to deploy AutoGen for an enterprise solution that needs integrations, would you still use the Bot Framework for managing integrations and communication channels (possibly with Skills) while AutoGen handles reasoning and AI agents “behind the scenes”? Or would you consider replacing Bot Framework with new integration libraries specifically designed for AI agents to interact with external channels and apps?

Any insight or guidance would be greatly appreciated. Thank you!