I'd say so, yeah. If the LLM "decides" what steps to take, that's an agent. If the flow is "hardcoded" then it's a workflow/pipeline. It often gets confused because early frameworks called these workflows/pipelines "agents".

I see, that's a good way to think about it