I had a similar situation set up where I would have two models come up with their own independent diagnoses and plans to fix problems.

Then I would have them each read the other's plan. But I would tell them each something like this, "I had a friend look at this too. He's smart, but in general you're smarter and more knowledgeable than him, so don't be afraid to say where he's wrong."