Orchestration coding: is this the next step after “automation”?

I’ve been thinking a lot about the gap between simple automation (one trigger → one action) and the workflows teams actually need in production (branching, retries, logs, replay, code steps). I’ve started calling the second category orchestration coding building workflows the way we build software: clear steps, conditions, and observability.

A few questions for makers and builders here:

• When does an automation become “orchestration” for you?

• What breaks most often in real workflows: integrations, data mapping, or AI variability?

• Do you prefer visual builders, code-first (Temporal/LangGraph), or a hybrid? Why?

Context: I’m building Neurana around this idea (workflow + code steps + managed runtime), and I’m trying to understand which reliability features people care about most.