A document pipeline always runs extract, validate, then store, in that order. Which design is the better fit?
Why C is the answer
A fixed workflow that calls Claude at each step
When the steps are known and do not depend on discovery, a workflow is cheaper, faster and easier to test. Agents earn their cost when the path is not known in advance.
In detail
- The steps are known and always run in the same order.
- A fixed workflow is cheaper, faster and easier to test.
- Claude can still do the work at each step.
- Use code for steps that need no judgment.
Use case
An illustrative scenario, not a real incident.
An insurance team always extracts, validates and stores each claim. A fixed workflow calls Claude for extraction, and plain code handles validation and storage, so each step can be tested alone.
See it in different ways
flowchart LR A["Extract"] --> B["Validate"] B --> C["Store"]
mindmap
root((Patterns))
Prompt chaining
Fixed steps
Routing
Classify first
Parallelization
Sectioning
Voting
Orchestrator workers
Subtasks found at run time
Evaluator optimizer
Feedback loopflowchart LR S0["Extract with Claude"] --> S1["Validate in code"] --> S2["Store in code"]
Key point: Use a workflow when the steps never change.
Why the other options are wrong
- A. An autonomous agent that decides the order itself
- An agent choosing the order adds risk with no benefit, since the order never changes.
- B. A manager agent supervising three subagents
- Three subagents add calls and coordination to a simple sequence.
- D. One long prompt that does all three steps at once
- One long prompt makes each step harder to validate on its own.