Before handing a task to AI, I want to clarify where it needs to stop.
While working on my website for Open Studio 17, I used AI to prepare and implement changes. I kept sensitive configuration decisions under my control, reviewed the work and challenged the suggestions. I brought my own thinking into what needed attention and why.
I asked for each proposal to be discussed before implementation, feedback to be collected before revisions, and the updated version to be checked before release. At each review, I could question the approach, request adjustments or decide it wasn’t ready.
I think five things need to be clear before delegating work to AI: 1. What are we trying to achieve, and what should stay unchanged? 2. What can AI work on within the agreed scope? 3. In what order should the steps happen? 4. Where must it stop for a human decision or approval? 5. What should happen when something is unclear or falls outside the brief?
A review point is useful when the person reviewing has enough context to question the output, request changes and decide whether to continue. That’s why I use AI to support both thinking and execution, while keeping judgment and responsibility with me.
For me, responsibility for defining the problem, choosing the approach and approving the final result stays with the person leading the work.