The framing of “AI engineering” as prompt engineering is a category error. Prompts are the easy part. The hard part is what happens around the prompt: identity, scope, gates, state, audit, recovery, retry, fan-out, fan-in, idempotency, rollback, post-incident learning.
I spend maybe five percent of my design time on prompts. The other ninety-five is the workflow. When the workflow is right, mediocre prompts produce excellent outcomes. When the workflow is wrong, perfect prompts produce reliable disappointment.
If you’re hiring for “AI engineering” and the interview is about prompts, you are interviewing for the wrong skill.