Chapter 1 · Practice 1 of 4
Start with the simplest design and add autonomy only when it pays off
Why it matters
Anthropic's article on building effective agents draws a clear line:
| Design | How it works | Best for |
|---|---|---|
| Workflow | Model calls and tools orchestrated through predefined code paths | Well-defined tasks with known steps |
| Agent | The model directs its own process and tool use | Open-ended problems where steps cannot be predicted |
The article's advice is to start with the simplest solution and add complexity only when needed. Agents often trade latency and cost for better task performance, and a single model call with retrieval and good examples is often enough.
Before you build an agent, write down:
- the steps a person would take to do the task;
- which of those steps are always the same;
- which steps genuinely need a decision that depends on earlier results.
If the answer to the third question is "none", build a workflow.
Avoid
- Building an autonomous agent for a task that always follows the same steps