Claude Certified Architect – Foundations Practice Question 1: Workflow vs Agent
Claude Certified Architect, Foundations · Test 1 · question 1 of 10
A team builds a feature that turns a support ticket into a customer reply. The steps are always the same: classify the ticket, retrieve the matching policy text, draft the reply, then check the draft against the policy. Each step has a clear input and output, and the team values predictable cost, latency, and easy debugging over flexibility. Which design is most appropriate?
Show answer and explanation
Answer:
B is correct. Anthropic's guidance distinguishes workflows (LLMs and tools orchestrated through predefined code paths) from agents (the LLM directs its own process). When a task decomposes into fixed, well-understood steps, a prompt chain gives predictability, lower latency and cost, and per-step debugging, and you can add programmatic checks between steps. Agents are for open-ended problems where the number and order of steps cannot be predicted.
A is wrong because autonomy adds cost, latency, and variance with no benefit when the path is known. C is wrong for the same reason, and dynamic subagent creation adds even more overhead. D is wrong because it gives up the per-step control, intermediate validation, and debuggability that the team wants; a single call that does everything is harder to inspect and tune.
Reference: Building Effective Agents