When two AI models give you opposite answers, how do you decide?

We are launching Polora here on July 14, but before that I wanted to actually talk to this community about the problem behind it.

For a while I would ask one AI model an important question, get one fluent, confident answer, and just go with it. No second opinion, no sense of where it might be wrong. Making calls I could not easily undo, that started to bother me.

So I got into the habit of putting the same question to a few different models. The useful part was never where they agreed. It was where they split. That disagreement usually pointed straight at the real risk, or the thing I had not thought about.

I would genuinely like to hear how other builders here handle this. When you get an important answer from an AI, do you cross-check it, and how? Have you ever caught a confident but wrong answer because another model disagreed?

Not pitching anything today. Just trying to learn how you all actually do this.