What would it take for an agent to handle a task from start to finish?

Start with the outcome you’d want from an independent agent. Work backward to see what it could handle today, what’s missing, and where it would still need you.

Full prompt

Read the first question in [Stella Garber’s “Is Your AI Idea Any Good?”](https://every.to/p/is-your-ai-idea-any-good-a-5-question-stress-test-from-a-serial-founder), where Hoop rethinks task management around capturing commitments automatically. Use that example to start with the outcome people need. Based on what you know about my work, pick one recurring job and describe the experience I’d want if an agent could handle it from start to finish. What would need to be true for that to work? Consider the context, access, capabilities, and ways to check the result it would need. Distinguish what is possible today from assumptions we would need to test, and explain where my judgment or approval would still matter. Suggest a useful first version I could try now and how we would know whether it worked. If you need context, ask one question. Finish with a short version of the thinking lens I can keep.