How to tackle big company bets with AI

If you have been asked to run a workshop on a big bet, the hard part is getting a group past what it already knows.

We designed the programme together with the team at a large Nordic company, and ran the first day with people from across the company working in groups, each with an AI agent tool. This guide is how we ran it, step by step, so you can run your own.

The dreamer room

Together with the company's team, we chose the Disney creative strategy as the frame for the series, described by Robert Dilts, an author and trainer who studied how Walt Disney worked. Dilts saw three roles in Disney's work, a dreamer, a realist and a critic.

This first day was only the dreamer room. Three rules held all day.

• Think big first. No cost, data or systems today, and no "we tried that".
• No criticism before there is a plan. Objections go on a parking lot.
• No defence of today's way of working. For the day, we treat it as something we could drop.

Four kinds of unknowns

We sorted what the group knows into a simple grid, adapted from the Johari window by the psychologists Joseph Luft and Harrington Ingham.

Most workshops stay in the top row. This day was about the bottom row.

Before the day

We set up one shared folder for the whole series, with the background, the interviews, the ideas and the assumptions. Every session reads from it, and what the group finds goes back into it at the end. It works like an agent in a folder, shared by the group.

Before the day, AI interviewed each participant to gather context, one question at a time. Each summary went into the folder, so the groups started the day with what people know but rarely write down.

YouCopy

Interview me about [the bet] before we start. Ask one question at a time and wait for my answer, at most ten. Ask especially about what I do from experience without thinking about it, what I take for granted and what I find hard or frustrating. When you have enough, write a short summary under the headings Situation, Decisions, Information, Friction and Open questions.

On the day, each group worked in an AI agent tool with its own copy of the folder, and the agent saved what it made straight into it.

Deep dive

The day started with one idea from the bet. Each group asked the agent to explain it as if they were new hires, and to mark what it was unsure of.

YouCopy

Read the files in the folder. Start from [one idea from the bet]. Explain to us, as if we were new hires, what happens, who does what and which decisions must be made. Mark what you are unsure of, and show where in the material you found it.

The agent's mistakes started the best conversations. When it got something wrong, someone in the group explained how it really works, and that went into the folder.

Remove the limits

Next, each group asked the agent what they had not thought of, and which assumptions came up again and again in the interviews.

YouCopy

You have read everything in the folder. What have we not thought of? List ten things we do not know that we do not know, and the questions someone from a completely different industry would ask us. Do not judge whether anything is feasible.

Then they took out one assumption at a time and asked what would happen if it were not true, from the angle of data, the market or the organisation.

YouCopy

Today we [describe how the work is done]. Which assumptions lie behind the way we do it? Split them into two groups, the ones that are probably true and the ones that are only habit. Then suggest how we could do it if we started from scratch today.

Each group kept two what-if questions for the afternoon.

Prototypes from the future

In the afternoon each group imagined that one of its what-if questions had come true in five years, and built something from that future. A clickable prototype makes an idea concrete enough to talk about, long before anyone knows how to build it.

YouCopy

Imagine that [our what-if question] is true in five years. Make a clickable prototype in HTML that shows what it could look like. It can be a tool, an idea or a new way of working. Make clear what is impossible today.

Each group made many rough drafts first, then picked two or three and made them clearer round by round, adding a few words to the prompt each time. The magic powers of HTML has the words that turn an answer into a page. When an idea felt too careful, they asked the agent to push it.

YouCopy

Here are our ideas from today [paste the ideas]. Imagine we look back on them in five years. Which of them will seem embarrassingly unambitious? What should we aim for instead?

Combine the ideas

Then each group swapped its best prototype with the group next to it and asked the agent to combine the two.

YouCopy

Here is our concept and the neighbouring group's [paste both]. Make five new concepts that combine them, without asking whether they are realistic yet. What becomes possible that neither can do alone? And how would a newcomer that wants to make us redundant do it?

Each group wrote, in its own words, what is impossible today but valuable. Whether it can be done is a question for the next session.

Showcase and save

Before the short showcase that ended the day, each group put its prototypes, the parking lot and the sticky notes in the folder.

YouCopy

Save what we found today in our insights file, and put what we did not get to in the parking lot.

In a classic workshop the sticky notes stay in the room. Here they become lines in a few files that the agent reads at the start of the next session. That is how each session builds on the last, as in building AI capital.

The next session starts with one line.

YouCopy

Read our notes from last time before we start. What do you remember?

Run your own in half a day

A half day on one bet is enough to try it.

• The week before. Let AI interview everyone who will take part, and put the summaries in one shared folder.
• Deep dive. Let each group ask the agent to explain one idea as if they were new hires.
• Remove the limits. Ask what you have not thought of, and keep two what-if questions.
• Prototypes. Make rough drafts first, then pick two or three and make them clearer.
• Save. Put everything in the folder before you leave.

We have more sessions planned. They start from what is impossible today but valuable, and from everything we saved in the first one. The realist room asks how to get there, and the critic room asks what must be stronger.