Reveal a problem to fresh minds one layer at a time, so their first assumptions can't box them in.
The moment people hear a problem, they start matching it to solutions they already know. Unfolding delays that moment. A facilitator who knows the real problem gathers a group who don't, and starts the conversation far away from it, with a broad, human theme.
After each round of ideas, the facilitator reveals one more layer of detail, moving the group closer to the real situation. Only at the end is the actual problem stated. By then the group has explored widely, and ideas from the early, open rounds can be carried down and adapted.
Use it when a problem is so familiar that everyone has fixed views, when experts keep proposing the same solutions, or when you want a fresh group's thinking without loading them with your own biases first.
Start wide and reveal the problem one layer at a time. Ideas from every layer flow down to the final question.
Choose a part of the diagram or a tab to explore each layer.
What big, human theme sits behind this problem?
What happens in the broader area your problem belongs to?
What's special about this kind of situation, and the people in it?
What changes when you add the real place, people and constraints?
Now that you know the real problem, which ideas from every layer could solve it?
Unfolding works by delaying the moment when people's existing assumptions lock in.
The first version of a problem people hear tends to anchor everything that follows: later ideas stay close to it. Starting broad means the group's first anchor is a wide theme, not a narrow problem with an obvious fix.
Ideas about a broad theme (what makes time fly, what helps a morning go well) can't be dismissed as unrealistic, because there's no reality yet. They become a stockpile of angles you can adapt later.
People who don't know the history don't know what “won't work”. Holding back the details protects their outsider view, which is exactly what you invited them for.
Moving down one level at a time lets the group test earlier ideas against new details. Many good solutions turn out to be an early idea, reshaped by a later constraint.
Before the session, the facilitator writes the actual problem in one sentence and keeps it private. It's where the group will arrive at the very end.
Work backward from the problem to 3–5 broader layers, ending with a wide, human theme. Each layer should add one meaningful piece of detail. Write one open question for each.
Invite 4–8 people who don't know the problem: colleagues from other teams, users, friends of the project. Tell them only that you'd value their creative help.
Open with the broadest question and generate ideas freely. Capture everything on the wall. Resist dropping hints about where it's heading.
Reveal the next layer, ask its question and generate again. Before moving on, ask which earlier ideas still apply or could be adapted. Keep every idea visible.
State the real problem. Ask the group to sweep the whole wall for ideas that fit, combine them, and shortlist the best with a quick NUF Test.
A facilitator's script with 3–5 layer questions, a long wall with one column per layer, sticky notes and an hour. With AI as an extra fresh mind, use the prompt below.
Subject: breakfast at a 1,200-student high school. Fresh group: six parents from a nearby district, a café owner and a soccer coach. Constraint: the cafeteria opens only 25 minutes before first bell.
“What helps anyone start the day well?” Ideas: routine, a friendly face, something to look forward to, no decisions to make.
“How do teenagers start their days?” Late nights, rushed mornings, no appetite at 7 a.m., and hating to look as if they need help.
“The 20 minutes before first period at a big school.” Crowded buses, friends at the gate, nobody walks to the far end. Idea: bring things to the buses.
“Food in those 20 minutes, with a cafeteria open for 25.” Ideas: grab-and-go bags, eating in homeroom, music and a friendly face at the door.
“How might more students eat breakfast?” Favorites: a cart at the bus drop-off, and a second breakfast after first period for late appetites.
62 ideas across five layers. Two of the three favorites grew from layers 2 and 3, before anyone knew the topic was breakfast.
Unfolding needs people who don't know the problem yet, which is hard to arrange and impossible to repeat. An AI can play that role on demand, as long as it hasn't been told the real problem. You stay the facilitator: you plan the layers, reveal them in order and decide which ideas carry forward.
Use one chat to design the layers, then a brand-new chat to run them.
My problem is X. Suggest 4 broader layers above it, from a universal theme down, with one question each.
In the fresh chat, give one layer at a time and never reveal the destination.
Give me 12 ideas for this question: what helps anyone start the day well? Don't guess what it's for.
After each new layer, ask which earlier ideas still fit or could be adapted.
Here's more detail. Which of your earlier ideas still apply, and how would you adapt them?
State the real problem and ask for a sweep of every idea from every layer.
The real problem is X. Go through every idea from every layer and propose the 5 best adaptations.
Describe the real problem, pick a mode, then copy the prompt into Claude, ChatGPT, Gemini or Copilot. In coach mode the AI plans the layers with you, then helps you run them round by round, adding outsider ideas without giving the problem away.
The “Open in” links put the prompt in the web address. If your challenge includes anything sensitive, copy and paste instead.
After the session, click one to copy it, then paste it into the same chat.