Good managers delegate tasks. Great managers delegate outcomes. This is true both when managing people and when managing agents.
There's a training exercise the US military uses where commanders are sent to an elementary school. They arrive thinking they're going to control every movement of every kid on that playground. Of course it doesn't work. The kindergarten teachers already know the answer: set boundaries and give the kids freedom to do whatever they want within them.
I learned this lesson long ago in my engineering management career. I'd review code and find the engineers approached it differently than I would have. But, in the end, it still got the job done in all the important ways, so I approved.
It's the same for agents: I define what success looks like and I let them implement it. My gates define the boundaries. Sometimes they are deterministic: does the plan document have the proper structure; does the code pass lint. I also use a different agent to do qualitative review: is it a rational plan? Does the code properly separate concerns?
The "great manager" approach: define the gates that do the managing, while you do the leading.
How do you draw the line between directing versus micromanaging your agents?
The framework behind this: Trust Topology →