Feature Request: Hierarchical Team Conversations in AutoGen Studio

Background

Currently, AutoGen Studio allows users to configure multiple agents and orchestrate their conversations in group chat settings. However, it lacks support for hierarchical agent structures where each representative agent can consult with an internal sub-team before providing a response in the outer conversation.

What I want to achieve

I have multiple teams, each consisting of:

• A domain specialist agent (e.g., Analyst, Product Owner, Marketing, QA, Architecture, Construction, UI/UX, Deployment).
• A human agent (UserProxyAgent) representing human input.

Each team has a representative agent that participates in a larger "representatives group chat."
When it's the representative’s turn, it should:

1. Consult with its internal sub-team (specialist agent + human agent).
2. Continue the sub-conversation until the human agent approves the result.
3. Summarize the approved outcome and share it in the outer group chat with the other representatives.

Why this matters

• This pattern models real-world organizational decision-making: representatives don’t act alone but rely on internal consultation.
• Enables human-in-the-loop governance while still leveraging LLM-powered specialists.
• Supports multi-layered conversations (inner team + outer representatives), which is crucial for enterprise-grade workflows.

Proposed solution

• Extend AutoGen Studio to allow nested group chats (sub-conversations).
• Add support for conditional looping where an agent must obtain explicit approval from a human agent before exiting the inner chat.
• Enable configuration of this flow via the Studio’s UI (or JSON definition).

Example use case

Teams:

• Analyst team
• Product owner team
• Marketing team
• QA team
• Architecture team
• Construction team
• UI/UX team
• Deployment team

Each team = { specialist agent + human agent }.
Representatives form a "representative council" chat.
Flow = representative → inner team consultation (until human approves) → outer council chat.

Benefits

• Provides greater modeling power for complex organizations.
• Helps researchers and product teams simulate decision-making hierarchies.
• Makes AutoGen Studio more aligned with real-world enterprise workflows.

References

Related concept in core AutoGen: SelectorGroupChat.
Enhancing AutoGen Studio to support this would bridge design and execution.