What leads to a better result, a good process or smarter people? When it comes to agentic systems, does a good workflow beat a more expensive model? I certainly learned a valuable lesson doing this experiment, and it's not the obvious one. Can you spot it?