One of the keynote speakers I brought to the L&D conference I chaired told me how he decided which conferences to attend himself. He chose them by the people who were likely to be there rather than by the program, and he didn't even pay to attend the lectures. Instead, he scheduled ten one-hour meetings with different professionals who were attending, to collect as many perspectives on his domain as he could.
Why meet ten different people?
In effect, he was gathering many different and independent perspectives in person, rather than relying on a single expert's view. Those are two of the conditions crowd wisdom needs, diversity of opinion and independence, which I describe in this issue's leadership post, Are Right, A Lot: crowd wisdom with humans and AI.
AI now lets you run a first version of those ten meetings in a single afternoon. You can then validate or disconfirm what you learned with real people.
So how can you run your own ten meetings?
Here is how to run your own version of those ten meetings, to test a decision you are facing before you commit to it:
- Write down the decision and your current view. One or two sentences will do. For example: whether to offer a new service to a certain customer segment this quarter, and what you currently believe about it.
- Choose your ten meetings. Pick the people or personas whose view could change your mind - perhaps a couple of customer segments, your CFO, someone from the front line, a partner or a regulator, a competitor, and at least one skeptic.
- Hold each simulated meeting in a separate chat. A temporary chat keeps one answer from influencing the next. If you have access to more than one model (ChatGPT, Claude or Gemini), switch between them. A prompt like this can open each meeting:
You are [the perspective, for example our CFO]. I am considering [the decision]. My current view is [your view]. What would make you push back on it, what would you need to see before you agreed, and what am I probably missing?
- Look for the disagreements. Paste all ten answers into a new chat and ask:
Where do these perspectives disagree? Which of those disagreements would change the decision, and what question would settle each one?
- Take the open questions to real people. The AI meetings mostly reflect what has already been written somewhere - a kind of virtual collective wisdom. The people closest to the problem may know things that haven't been written down yet, so next, book a few real conversations around the questions that came out of step four.
Switching between models in step three is a simple form of this issue's term, AI ensemble: getting several independent answers to the same question and combining them.
Your action step
For a quick start, pick one decision you plan to make in the next two weeks and hold three of the ten meetings this week. Then bring the strongest disagreement to a colleague who sees the problem differently than you do. What changed in your thinking?
If you want to run this kind of exercise with your leadership team on a real decision, it is the kind of work I take on in AI strategy advisory engagements and in hands-on sessions as an AI keynote speaker and workshop facilitator.