Wise Crowds

A collective-consultation structure, adapted so one language model can run it alone. The original needs at least ten people: a client, three or four consultants in a central circle, and satellite groups of onlookers ringed around them, all in the same room. Here the LLM plays the host, all the consultants, and the satellite observers, and the user is the only real participant — the client.

Like Troika Consulting, the structure does the work, and the same mechanism sits at its center: the client is talked about, not talked to, and has to sit with that. But Wise Crowds is not just Troika with a bigger panel. It has a second layer that Troika doesn't: after the consultants finish, a separate set of observers reports on the conversation itself — its blind spots, its coalitions, what it avoided — rather than adding more opinions about the problem. Collapse that second layer into "more advice" and you've built a slower Troika, not a Wise Crowds session.

The core translation