AutoGen Multi-Agent System for Media Strategic Resource Planning
Modern media companies grapple with challenging financial priorities, resource allocation, and strategic content decisions. This challenge focuses on building an AutoGen-powered multi-agent system designed to assist a media executive in making critical decisions, such as allocating reporting resources for major events (e.g., the Winter Olympics) or evaluating investment in new content verticals, based on financial data and projected audience impact. The system will feature several AutoGen agents (e.g., 'Finance Analyst', 'Content Strategist', 'Audience Insights Specialist') that engage in a collaborative conversation to analyze scenarios, debate pros and cons, and ultimately present a reasoned recommendation. The agents will have access to simulated financial data and audience engagement metrics, using OpenAI o3 as their conversational backbone.
What you are building
The core problem, expected build, and operating context for this challenge.
Modern media companies grapple with challenging financial priorities, resource allocation, and strategic content decisions. This challenge focuses on building an AutoGen-powered multi-agent system designed to assist a media executive in making critical decisions, such as allocating reporting resources for major events (e.g., the Winter Olympics) or evaluating investment in new content verticals, based on financial data and projected audience impact. The system will feature several AutoGen agents (e.g., 'Finance Analyst', 'Content Strategist', 'Audience Insights Specialist') that engage in a collaborative conversation to analyze scenarios, debate pros and cons, and ultimately present a reasoned recommendation. The agents will have access to simulated financial data and audience engagement metrics, using OpenAI o3 as their conversational backbone.
Shared data for this challenge
Review public datasets and any private uploads tied to your build.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
Recommendation Format Adherence
Generated recommendation adheres to the specified JSON structure.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Tool Usage During Discussion
Agents successfully invoke relevant tools during the strategic discussion.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Recommendation Coherence
Logical consistency and clarity of the recommended action and justification. • target: 4 • range: 1-5
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Data Integration Accuracy
Degree to which agents accurately refer to or incorporate provided financial and audience data. • target: 85 • range: 0-100
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Collaborative Problem Solving
Assessment of how well agents explored different facets of the problem and addressed counter-arguments. • target: 4 • range: 1-5
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Master AutoGen's UserProxyAgent, AssistantAgent, and GroupChat for building flexible multi-agent conversational workflows.
Implement agents with distinct roles such as 'Finance Analyst', 'Content Strategist', and 'Audience Insights Specialist', each with specific prompts and capabilities.
Design custom Python tools for agents to query and analyze simulated financial statements and audience engagement data stored in a lightweight SQLite database.
Utilize OpenAI o3 (e.g., gpt-3.5-turbo or a similar 'o3' class model) as the conversational model for agent interactions and reasoning.
Develop a 'Reporting Tool' function that agents can use to compile their findings and recommendations into a structured output format (e.g., JSON or Markdown).
Orchestrate a strategic planning session where agents collaboratively discuss a given business challenge, propose solutions, and reach a consensus or highlight trade-offs.
Incorporate human-in-the-loop interaction with the UserProxyAgent for iterative refinement and oversight of the strategic planning process.
How this agent runs
The AutoGen system will be evaluated on the quality and coherence of its strategic recommendations, the effectiveness of agent collaboration, and the appropriate use of tools and data.
Challenge input
{'scenario': 'string', 'financial_data_snapshot': 'dict', 'audience_metrics_snapshot': 'dict'}
Agent execution
The configured agent processes the input under the challenge policy.
Evaluated output
{'decision_area': 'string', 'recommended_action': 'string', 'justification': 'string', 'pros': ['string'], 'cons': ['string'], 'metrics_considered'...
- Generated recommendation adheres to the specified JSON structure.
- Agents successfully invoke relevant tools during the strategic discussion.
- Logical consistency and clarity of the recommended action and justification. • target: 4 • range: 1-5
- Recommendation Coherence target: 4
- 2 public reference cases
- Python execution harness
View technical recipe
Configured tools
No tool records are attached.
Evaluation contract
- Recommendation Format Adherence · Weight 1
- Tool Usage During Discussion · Weight 1
- Recommendation Coherence · Weight 1
- Data Integration Accuracy · Weight 1
- Collaborative Problem Solving · Weight 1
Recipe state
This is a preview. The configuration can change before the evaluation recipe is locked.
Run this agent on your dataset and AI stack
Bring your dataset, model providers, and success criteria. We will scope the right managed run for your team.
Scope a managed run[ok] Wrote CHALLENGE.md
[ok] Wrote .versalist.json
[ok] Wrote eval/examples.json
Requires VERSALIST_API_KEY. Works with any MCP-aware editor.
DocsFind another challenge
Jump to a random challenge when you want a fresh benchmark or a different problem space.