When is a fixed workflow better than a more autonomous agent?
Answer out loud or in your head first, then compare.
Answer
When I know the steps and conditions, I usually choose a fixed workflow. For example, a document has to be read, its fields checked, sent for approval and saved. A model can help in one stage without choosing the whole path itself.
More freedom is useful when I don't know in advance how to reach the goal, e.g. when exploring an unfamiliar repo. The cost is harder testing, more possible failures and higher cost. You can combine the approaches: give the model a choice in a limited place and control the rest with code. First I check whether the extra autonomy actually improves the result over a simpler solution.
Why it matters: A basic design decision for agentic systems.
Possible follow-ups: Which decisions should never depend solely on the model's suggestion?
How did it go? Be honest — “Not yet” puts the question on your review list.
Start with
- H-P01 · How does an AI agent differ from a plain model call? — Agents and workflows
- H-P02 · What is a workflow, i.e. a fixed sequence of work? — Agents and workflows
Further reading
- Anthropic: Building effective agents(opens in a new tab)19.12.2024. Workflows versus agents; architectural patterns. The tools in the article have evolved since.