A good prompt is not a workflow. What turns an AI task into a repeatable process?
retained text
I’ve been thinking about where the line is between “using AI well” and actually changing how work operates. A prompt can work beautifully, but if a person still has to remember to run it, collect the inputs, inspect the result, route it somewhere, and decide what happens next, then we’ve automated one cognitive step — not the process. For recurring work, I now think you need at least five things: \- a trigger \- authoritative inputs / source of truth \- acceptance or verification criteria \- an exception path \- a clear human decision boundary Simple example: instead of a manager asking every Monday “summarize project status,” the process should notice the relevant changes, prepare a brief from approved sources, cite the evidence, and surface only the issues that actually require a decision. I’m curious from people running agents in real work: which part is hardest in practice — triggering/context, verification, exception handling, or deciding when a human must take over? My suspicion is that the hard part is increasingly not the prompt. It’s the process design around it.
document fetched 2026-09-15 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1wg48o4/a_good_prompt_is_not_a_workflow_what_turns_an_ai