Support messages fall into billing, technical and refund types, and each needs different instructions and tools. Which pattern fits?
Why B is the answer
Routing
Routing classifies an input and sends it to a specialized path. It keeps each prompt focused, because instructions for one type do not have to share space with the others.
In detail
- The inputs fall into known types, and each needs its own handling.
- Routing classifies first, so each prompt stays focused.
- Add an "other" path, so unclear messages still get handled.
Use case
An illustrative scenario, not a real incident.
A support team has billing, technical and refund requests. A small classifier sends each to its own prompt and tools, and a change to refunds cannot harm billing.
See it in different ways
flowchart LR
A["Input"] --> B{"Router"}
B -->|"Billing"| C["Billing prompt"]
B -->|"Technical"| D["Technical prompt"]
B -->|"Refund"| E["Refund prompt"]mindmap
root((Patterns))
Prompt chaining
Fixed steps
Routing
Classify first
Parallelization
Sectioning
Voting
Orchestrator workers
Subtasks found at run time
Evaluator optimizer
Feedback loopflowchart TD
Q{"What type of message?"}
Q -->|"Billing"| A0["Billing path"]
Q -->|"Technical"| A1["Technical path"]
Q -->|"Refund"| A2["Refund path"]Key point: Routing keeps each prompt focused.
Why the other options are wrong
- A. An autonomous agent
- An agent is more than this task needs, since the types are known.
- C. Evaluator-optimizer
- Evaluator-optimizer improves one output through feedback.
- D. Prompt chaining
- Chaining runs fixed steps in order, and it does not choose between paths.