Is an agent's plan something to execute without changes?
Review note: The answer refers to the original author's CV or target job offer — adapt it to your own experience.
Answer out loud or in your head first, then compare.
Answer
A plan is a proposed way of working. After reading the code or a tool result, it may turn out it needs changing. That doesn't automatically change the goal, criteria and permissions. The application or a person sets those boundaries.
Example: an agent plans to fix a small function but proposes replacing a whole library. That's a significant change of scope and needs assessment. For a known process, pre-written steps can be more predictable. Ticking off a plan doesn't prove the work was done either: you need a test result or a record in the system. In a short interview, the plan should help you act, not take up half the time.
Why it matters: Explains the practical meaning of planning from the offer.
Possible follow-ups: When does new information justify changing the plan, and when is the agent just drifting from the task?
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.