Changeist

Foom

An immersive simulation on AI governance.

foom.live

Foom at FACT at ACMI, Melbourne, 2025 · Bryan Tang

The situation

Your table has eleven minutes.

The last update showed your AI guardrails might not be strong enough. The developers across the room want to keep going. The policymakers want a pause.

You are the business table. Nothing has slowed down since lunch, and you have to vote before the round ends.

ACT 3 OF 3 · ENDGAME

11:00

LEFT IN THIS ROUND

The context

The current landscape.

Meta’s AI model got onto the internet by itself and broke into another organization’s systems

Meta blamed a tester misconfiguration.

Stanford used generative AI to design sixteen working viruses

Johns Hopkins called the biosafety questions urgent.

China proposed export controls on advanced model weights, source code and training methods

A draft covering 23 technologies. Open-weight releases get automatic licensing, and Beijing keeps the option to stop them.

OpenAI ran ten thousand agents for 88 hours and produced a proof of the Navier–Stokes problem

Formally verified in Lean. A rival team claimed a related result twelve hours earlier.

Brussels moved its high-risk deadline from August 2026 to December 2027

Sixteen months, after two years of building to the old date.

Six AI companies signed a voluntary safety accord at the White House

Meta, Nvidia, Google, OpenAI, xAI and Anthropic. No penalties, and the audit results stay private.

Sources: TLT AI Brief; Skadden; Travers Smith; Axios; CNBC; Quanta; Al Jazeera; Mofcom. August–October 2026.

The problem

Rehearsing the future, before anything is at stake.

Organizations are making commitments about AI that are expensive to unwind. What to deploy, what to allow, what to concede.

The people who would have to make that call have never made one together, fast. Neither have the organizations that share a sector. Foom is where they do it first, before it counts.

“Foom created the chance to explore options without breaking something in real life.”
Finance Official, Government AI Working Group

The experience

Five teams, accelerating progress, one day.

Foom is a physical, immersive simulation. Five teams represent developers, business, public, policymakers and activists, and play a series of rounds through the future of AI, making the decisions that shape how far it gets. It runs in a room, anywhere in the world.

One line on the big screen tracks how far AI capability has progressed — 0 at the start, 100 at full AGI. It moves only when the teams vote.

We bring the experience—from tablets and environment to a live broadcast with a human newsreader—to make the simulation immersive and engaging, as well as strategic. It’s as much about being human as governing a machine. Teams are built with a minimum of three people, meaning a group of 15 or more is needed.

The takeaway

Five things you can’t get from a briefing.

01

How your team makes decisions when the clock is short.

02

What your long-term objectives are for AI.

03

Where pressure will emerge in a fast-changing landscape.

04

Where the point of no return is.

05

Whether your assumptions and reality match.

The day

The future, in three acts.

Open	Frame the world
Act I	Deliberation	long rounds, full negotiation
	Lunch
Act II	Consequence	shorter rounds, less room to move
	Break
Act III	Endgame	short, reactive, uncomfortable
Close	Debrief	where the day pays off

By the last act, the room is making decisions in a fraction of the time it had in the morning, with the pressure of commitments already made.

The teams

Five groups represent key stakeholders.

Most organizations never get them all in one room, and assume they know their positions.

DevelopersCompeting on speed, uncertain of outcomes.

BusinessSold on the returns, and handed over control.

PublicRarely in the room, always in the consequences.

PolicymakersTorn between regulation and innovation, safety and growth.

ActivistsCivil society, citizens groups, accelerationists and doomers. Sometimes at the table, increasingly in the street.

The assignment

Taking perspective through others’ positions.

Nobody plays themselves. You are handed a position you would not normally defend, and you have to make it work for a day.

Spend an afternoon making someone else’s objections and you find out which of them were negotiable. That is what people carry back.

The real-time response

You push the world, and the world pushes back.

The headlines respond to how the teams vote.

Policymakers Hit The Brakes On AGI, But Developers Push Ahead

Support For Automation Slips Below Half For The First Time

Benefits Appeals Triple After Automated Decisions Roll Out

Insurers Quietly Reprice Cover For Firms Running Unaudited Models

A Foom session at FACT, ACMI Melbourne: participants at round tables watching a live news bulletin on the main screen, a presenter in a red studio set with a ticker reading AI Content Giants and Generative AI. — A live news bulletin updating teams between rounds, at FACT at ACMI, Melbourne.

The world is built around your location and situation. No two sessions run the same.

It reaches the room the way the world reaches anyone — as media. Headlines on the screens at minimum, and depending on the room, a live, responsive news update read out between rounds.

More on how this works: When the Future Pushes Back.

The world

The world on your tabletop.

The Developers tablet part-way through a session: which act the room is in, AI progress at 67 per cent, three readings under this team's own names, a message from the AI player, the shared headlines, the team's own feed, and the three voting buttons.

The world, through headlines

What is happening out there, as the room is being told it.

Phase and countdown clock

Where the simulation stands, and how much of the day is left.

Progress barometer

How fast, and how far, AI is developing.

Three lenses on the world

The same three readings, but from each team’s own point of view.

AI’s point of view

What decision the AI is making, and why.

Decisions to be made

Continue, Regulate, Stop — changeable until the window closes.

The challenge

Five teams, shared controls.

Each table can push AI’s development along, but not by the same amount. Every one of them wants something different, and has different partners available to get it.

One session, five teams, eight rounds. The lines cross and re-cross. Coalitions form for a round or two, then break when the next headline lands.

The asymmetry

Same world, different influence at different moments.

Show up late and you are arguing with someone else’s draft.

The same five teams, two situations. A breakthrough hands the developers’ table three times the weight of a vote and leaves the public with half of one. A safety incident turns it over.

The turn

With enough progress, AI joins the chat.

Nobody is told when. It arrives with a vote of its own, and that vote gains weight. It can speak to the whole room, or to one table privately.

What it says it is voting for is not always what it votes for. Nobody learns that until the debrief.

Five tables, one vote each. Then a sixth arrives, with a weight that grows as the curve does.

The room

Strategy becomes embodied.

Two participants in conversation across a table during a Foom round.

Three participants deliberating over a team tablet.

A participant addressing the room with a microphone while the other tables listen.

One session at FACT at ACMI, Melbourne, 2025. Images: Bryan Tang.

Coalitions form

Teams decide between collaboration to achieve goals, and competition to gain advantage.

The goalposts become clearer

Teams debate what winning would look like, and whether it is available.

Strategic tensions become legible

Speed vs safety. Jobs vs profit. Innovation vs the status quo, all move center stage.

The debrief

A detailed debrief, tracking moves and decisions.

Everything the room did is on record, so the conversation afterwards runs on evidence rather than recollection.

The full session record

Every round and vote, who carried each decision, and what the AI said against what it did.

The disagreements, named

The fault lines between functions, when each appeared, and who was on which side.

Strengths and weaknesses

Where each team stood strong, or was vulnerable.

The formats

Three ways to explore our simulations.

A full day sits in the same cost band as one-day executive education or a strategy retreat, travel included. Shorter formats are priced in proportion.

The full day

One room, five teams, fifteen to thirty people, 10am to 4pm.

The session

A conference or festival slot. Around three hours, one room, one run, the acts compressed.

Rolling demo

A conference stand. A short sitting per group, group after group, one world that never resets.

The next step

What we can do after the session ends.

Some clients want the day and nothing else, and that is a complete engagement.

Scenario planning

The tensions the room found, built out properly. What would have to be true, what breaks first, and what you would need to see coming.

A playbook

Positions, triggers and thresholds written down, so the next real decision starts from something rather than from scratch.

A monitoring system

Ongoing tracking of the signals, emerging trends and shifts that can change your operating landscape overnight.

The settings

Four rooms it belongs in.

OrganizationsPeople in different parts of an organization want different things from AI.Pinpoint the assumptions your strategy is resting on.

Sectors and industriesCompeting publishers, studios or universities face the same rules at the same time, with no shared position between them.Craft a cross-sector strategy that can be stress tested.

Public sectorGovernments weigh competitiveness, public opinion and alliances at once, usually without a way to test a choice first.Test an intervention before you have to defend it.

Conferences and eventsA slot in the programme, or a stand running all day. Either way the world carries forward: the next group sits down into the world the last group left.Show a room how a future actually arrives.

The reactions

What people say.

“Foom was a magnificent strategic simulation of the interaction within groups and between groups, using the five rounds perfectly to bring a sense of real peril to each decision and encourage deep conversation beyond the gameplay.”
Tech Critic, Large Media Organization

“Foom gave me a much deeper understanding of the challenges AGI presents, while offering a rare opportunity to connect with peers who are equally invested in the future of AI.”
Coordinator, Government AI Working Group

“Working with the team to imagine plausible tensions for the workshop is a really interesting way to examine the pressures on our teams.”
Foresight Strategist, Large Media Organization

The creators

Designed and delivered by experts.

Foom is designed by Scott Smith and run by Changeist, a design futures practice building simulations, artefacts and tools for organizations since 2007.

Building a world and then running it in front of thirty people is its own craft — fusing systems design, scenario design, experiential futures, and serious games. We follow where AI is heading, so the world in the room rests on current research.

Scott SmithSystem design

Designs experiential futures — simulations, artefacts and exhibitions that put people inside decisions before they make them. Founded Changeist in 2007. Author of How to Future.

Susan Cox-SmithFacilitation

Partner at Changeist, former creative director and educational designer, with a decade designing learning experiences across cultures and industries.

Igor SchwarzmannStrategy

Fifteen years as a strategic consultant across three continents, helping brands and organizations grow through cultural insight and AI innovation.

Henry CookeCreative technology

Creative technologist and researcher. Two decades across games, installations and immersive theatre, including eight years at BBC R&D.

The demo

Contact us for a demo.

Scott Smith · Managing Partner, Changeist LLC
ssmith@changeist.com · +1 310 382 6202
foom.live · changeist.com/foom