Implement generative AI and agentic solutions 2.3 Optimize and operationalize generative AI systems

A contract-review app needs legal, financial, and compliance agents to each analyze the…

Study guide skill: Orchestrate multiple models, flows, or hybrid LLM and rules engines

Question B4-030

A contract-review app needs legal, financial, and compliance agents to each analyze the same contract independently, and then combine their findings into one report. The three analyses don't depend on each other. Which Microsoft Agent Framework orchestration fits best?

Select one answer.

  • A Group chat orchestration
  • B Sequential orchestration
  • C Magentic orchestration
  • D Concurrent orchestration ✓ Correct answer

Answer: D

Concurrent orchestration runs multiple agents on the same task in parallel; each processes the input independently, and the results are collected and aggregated. Magentic is for open-ended tasks where a manager agent plans because the solution path isn't known in advance. Group chat has agents collaborate in a shared conversation, and sequential runs them one after another.

Deep dive

How it works

Agent Framework's built-in orchestrations each fit a different shape of work:

  • Sequential: agents execute one after another in a defined order.
  • Concurrent: agents execute in parallel.
  • Handoff: agents transfer control to each other based on context.
  • Group chat: agents collaborate in a shared conversation.
  • Magentic: a manager agent dynamically coordinates specialized agents.

Concurrent orchestration lets multiple agents work on the same task in parallel. Each agent processes the input independently, and their results are collected and aggregated. Microsoft suggests it wherever diverse perspectives are valuable, such as brainstorming or ensemble reasoning. By default only the aggregator's output is the workflow's terminal output, and you can plug in a custom aggregator, for example one that merges the three findings into one report.

Three independent analyses of the same contract, combined at the end, is that pattern exactly.

The other options, and when you'd use them

  • Group chat: Agents collaborate in a shared conversation, building on each other. The analyses here must stay independent.
  • Sequential: Agents run one after another in a defined order. These analyses don't depend on each other, so nothing needs that order.
  • Magentic: A manager agent dynamically decides who acts next, for complex, open-ended tasks. Here the steps are known in advance.

Exam tip

Same input, independent parallel work, results merged = concurrent. Unknown path with a planner = magentic.

References

Last reviewed:

Ready to test yourself? The quiz asks this and every other AI-103 question, with instant feedback. Practice this in a quiz →

All Generative AI & agents questions · AI-103 study guide

Thanks for your feedback!