Concept guide · Agents, tools, and MCP
When is a multi-agent system better than a single agent?
A slower explanation for when the short interview answer is not enough. Start with the intuition, then connect it to the technical idea and the way an interviewer expects you to reason about it.
ML area
Generative AI → Multi-agent systems
Interview format
Advanced · Scenario
Question
GENAI-084
Start with the idea
In simple terms
Multiple agents are useful when work can be separated by expertise, tools, permissions, or parallel tasks. They are not automatically better: coordination adds latency, cost, and places where information can be lost.
Make it concrete
A research workflow might use separate retrieval workers for finance and product evidence, then a supervisor merges their source-backed findings before writing.
Build the answer
How to approach it
- 1Identify independent work or genuinely different capabilities.
- 2Define contracts and provenance for every handoff.
- 3Prove that coordination improves the outcome enough to justify its cost.
Go one level deeper
The technical explanation
Choose a pattern, supervisor and specialists, sequential handoffs, parallel workers, or hierarchical teams, based on the task graph. Define each agent’s contract, input and output schema, context boundary, permissions, and failure policy. Compare against one well-designed agent using task success, path efficiency, cost, latency, and cross-agent consistency.
Say this in the interview
Use multiple agents when tasks need genuinely different tools, permissions, expertise, or parallel context, and the coordination cost is justified by measured quality or throughput. A supervisor, sequential handoff, parallel specialists, and team-of-teams designs have different failure and observability trade-offs.
Keep preparing