Three Ways of Asking AI

While writing the book I ran a small experiment. I asked AI the same question in three different ways: “How should we govern AI systems responsibly?” What changed was not the machine. What changed was the kind of conversation I allowed.

1. AI as answer engine

This is how most of us use AI chats. I asked, and I received a well-structured overview of ethical principles, regulations and best practices. It gave me clarity and orientation. It did not give me any tension, reflection or meaning. Useful, but shallow by design.

2. AI as analytical partner

Next I asked the agents to improve the question before answering it: “Reframe this. What am I assuming?” Back came sharper questions. Who defines responsibility? At what scale does governance fail? Whose interests are protected, and whose are not? Better thinking, but it still stayed outside of me.

3. AI as dialogue partner

Finally I spoke with a persona grounded in moral philosophy and historical responsibility. It did not answer. It asked back: what do you mean by responsibly? Harm to whom? And then: what would you personally refuse to build, even if it were legal, profitable and admired?

That question landed. It moved the conversation from frameworks to my own position. Being questioned back created a depth that no better answer could have.

For me this is the core lesson. When we treat intelligence as a shared process, it becomes less about output and more about orientation. AI becomes a mirror for our own thinking, and the depth emerges between the perspectives.