I think it makes better decisions than me provided I give it enough high-level direction and context.

Sometimes I give it __too much__ direction and it finds the solution I had in mind but not the best.

I'm not into it enough that I'm formally running different personas against each other in a co-operative system but I kind of informally do that.