The mandate
to imagine.

Cities have no mandate to imagine the future

Thinking about the future is, almost nowhere, in a planner’s job description. Planning law is built to administer — check, zone, permit, archive — not to test and choose between futures. “The urgent eclipses the important,” trapping cities in cycles of reaction.

The warning sign: Ireland now lists town planning as an official shortage occupation — 62% of authorities struggle to recruit and retain planners.

Yet the pressure to transform has never been higher

Paris, the New Urban Agenda, the EU Green Deal, the European Climate Law, the Mission on Adaptation. National adaptation strategies rose from 20 to 32 countries in a decade; Covenant of Mayors signatories grew from a few hundred to over 4,500.

The mandate to deliver the future keeps growing. The mandate to imagine it does not.

Fracture 02 · UrgencyA decade of escalating commitments, 2013–2023. Source: European Environment Agency.

And the shared capacity to act together is thinning

In society, algorithmic curation splits the public into separate informational worlds and trust erodes. Inside the municipality, departments run in silos, budgets tighten, and staff are too busy operating the present to imagine the future.

Agents that synthesise, so people don’t transcribe

The hours we save are the manual ones: agents cluster, theme and summarise thousands of contributions. Officials review the synthesis instead of compiling it — the judgment stays human, the clerical work doesn’t.

Agents represent the absent — never the public

We tune agents to stand in for the people who can’t be in the room — children, future residents, the species a plan overlooks — and to stress-test scenarios. The living public’s voice is the one thing we will not automate away.

A plan becomes options people can react to

A site plan turns into explorable, supervised scenarios — not one hero render to admire, but variations citizens and professionals can compare, contest and rate. The image becomes an instrument for a decision, not decoration for one already made.

Deliberation at the scale of a whole population

The same generative loop runs online, between in-person sessions — reaching thousands who would never attend an evening workshop. Not a comment box: people propose, see consequences and converge, asynchronously, at city scale.

A documented mandate, not a comment count

Because every participant worked the same images and votes, the output is a ranked, visual, deliberated record of what a community can agree to want — defensible to a council, bankable to a developer, auditable later.

Policy pre-evaluation, while change is still cheap

Agents score each scenario against EU policy, the local design guide and city strategy as it is made — surfacing conflicts early, when an image is easy to change, instead of as a veto at the end of an expensive process.

Three moves, one method

Critique — fully automated processes optimise the existing mandate and inherit every blind spot of the system they accelerate.

Repair — restore what planning lost on its way to proceduralism: seeing consequences before deciding, hearing the voices the process excludes.

Innovate — a projective instrument: futures made visible, debatable and choosable.

Where the three roles meet: coplanning

Stakeholders, institutions and AI agents each bring something different. Coplanning is the overlap — and AI never holds the pen alone: every generative step is wrapped in a human step. Propose, then judge; render, then decide.

The output of a workshop is not a design. It is a mandate — documented evidence of what a community can agree to want, which professionals then translate into technical design.

1 · Sense

January 2023. A photo of a real street — the only input citizens need to start.

1 · Sense · Jan 2023Kasarmitori area street in winter, Helsinki.

2 · Decide

A citizen committee and entrepreneurs generate, debate and vote — in one room, around one table.

2 · Decide · WorkshopCitizens and planners co-designing around one table.

3 · Act

June 2023. The street as built — Helsinki Summer Streets, with the City of Helsinki.

3 · Act · Jun 2023The built summer street — parklets and greenery.

The sequence we all know too well

The ballot-box model is the wrong template for design: you choose once, blind, and live with it. Participation, then discontent, then demolition — conflict surfaces after the money is spent.

The pattern we build against

The Hippocratic Oath of working with AI

Bringing AI into a planning organisation is not a tooling decision — it transforms the institution. So the institution owes itself an explicit ethic. We ask every partner organisation to write their own.

The oath

Four commitments

01Augment judgment; never automate it away.

02Preserve the right to know consequences and change your choice.

03Make the model’s biases legible before asking anyone to trust its images.

04Represent those who cannot be in the room.

From commentators to placemakers

Children 6–14. With an Architects’ Council of Europe research grant we built a child-specific participatory AI — drawing-guided for the youngest, tablet co-design for the oldest — piloted in Finnish schools with Tampere University and at Milan Digital Week.

Children · 6–14Analog ideation → digital ideation → deliberation.

Exploring climate futures with the young

UNDP Accelerator Labs Panama ran workshop series with several hundred participants from local high schools and universities — turning photographs of their own waterfront into climate-adaptation scenarios they could debate and own.

Youth · ClimatePanama City — climate-adaptation scenarios by youth workshops.

Parameters beyond people

People tend to think people-first — or people-only. At BiodiverseCity? (Estonian Museum of Natural History, Tallinn), a tuned model let children explore what European greening policies mean: communal areas reimagined as wilder, more biodiverse habitats.

More-than-humanBiodiverseCity? — Tallinn, European Green Capital 2023.

Not a render to admire — a proposal to examine

Drag the handle. The site as participants photographed it — and as the youth groups proposed it: a proposal to examine, contest and own.

Panama City · UNDP Accelerator Labs

Current photographAI scenario↔

The honest caveat — customisation is also a new bias

Every tuned agent over-represents some features and under-represents others. Choosing a bias is choosing whose city gets imagined more easily — so we treat bias curation as design work, done in the open.

Bias, curated in the open

Four practices

DeclareEvery agent ships with a note on what it favours and what it flattens.

LocaliseTuned with local imagery and policy — with the client, not for them.

Counter-balanceReparative biases — children, elders, biodiversity — as open policy choices.

SuperviseProfessionals review against regulation and quality before decision.

Scored against EU policy and the local design guide

Bologna Verde, 2026 — with the Comune di Bologna, citizen scenarios for greener streets are evaluated by agents trained on EU policy and Bologna’s own catalogue of permeable, vegetal and mineral materials — checked while imagination is still cheap to change.

Bologna Verde · 2026A citizen scenario — greenery and a red cycle lane.

The design guide becomes evaluation criteria

Architectural quality enters the loop as data, not as a veto at the end. The city’s own catalogue of materials and solutions is what the agents score against.

The guide as dataTiles from Bologna’s nature-based-solutions design guide.

Testing a city model block by block

For the City of Vienna’s Supergrätzl programme, scenario evaluation helps planners and residents see how the model lands on specific streets — cooler, calmer, more playable — before commitment. Also in evaluation practice: Dubai Municipality, with planning law in the loop.

Vienna Supergrätzl · 2026The superblock model, tested in the city’s own visual language.

The same street, twice

Bologna Verde — the street as residents photographed it, and as the workshop reworked it, then scored against EU policy and the local design guide.

Bologna Verde

Street as foundCo-designed scenario↔

Trust as infrastructure

Munich · Zamanand Festival. 3,000+ ideas submitted, ~300 votes cast on reimagining the public space around the Siegestor — a festival-led experiment in rebuilding trust through social imagination.

Munich · GermanyThe Siegestor, reimagined — 3,000+ ideas, ~300 votes.

Residents design a future park

Dubai Municipality · UAE. Co-planning sessions where residents shaped an interactive play area and shaded cycling track — alongside our compliance-aware generation work with planning law in the loop.

Dubai · UAEDubai Municipality co-planning session.

A city-centre vision, co-planned

Lahti · Finland. A co-planning process feeding directly into the municipal strategy document — collective imagination given formal standing in the city’s most official text.

Lahti · FinlandThe city centre by the lake — vision feeding strategy.

Co-designing a capital city

Nusantara · Indonesia. Local communities co-designing the future capital — participatory AI at the most consequential scale there is: a city that does not exist yet.

Nusantara · IndonesiaA green low-rise neighbourhood for the new capital.

Regenerating Biloku, together

Pristina · Kosovo. A UNDP-led regeneration of a housing complex’s public spaces, with the mayor’s office at the table — participatory AI in a post-conflict planning context.

Pristina · KosovoResidents reviewing design boards at the Biloku workshop.

Reimagining vacant spaces

East Cleveland · USA. Seventh Hill’s Connect East Cleveland initiative — open-house co-design for vacant lots in a US disinvestment context, community-led from the first photograph.

Also: Helsinki (EU Best Case Award) · Zürich · Berlin · Tallinn · World Government Summit · HafenCity University Hamburg.

East Cleveland · USAOpen-house co-design with tablets.

The same conclusion, reached independently

Across development banks, the OECD and city networks, the same finding keeps surfacing: put the tools in the hands of the people who run the service, and use AI to widen participation — not to replace it.

Independent signals

The hard problems agencies bring us

One process that holds every stakeholder — deep in the room, wide across the population. Compliance and design quality carried as data. Participation that means changing minds in public, not submitting opinions. And speed that comes from front-loading agreement, not faster rubber-stamps.

Challenges

Four hard problems

One processA single workshop holds every stakeholder — in person and online at once.

Compliant & well-designedAgents tuned on regulation and design guides; accountable humans decide.

MeaningfulPeople propose, see consequences and converge — not a comment box.

Fast, honestlyThe expensive conflict moves to the cheap end of the process.

The openings participatory AI creates

Each engagement leaves a capability behind, not a report: a configured process the office re-runs freely, reach beyond the weekday-evening room, departments aligned around one picture, evidence both sides can stand behind — and a standing capacity to imagine in public, down to an expo kiosk the public walks into.

Opportunities

Six openings

A standard you ownConfigure once, re-run freely.

Reach everyoneCommuters, parents, shift-workers — online, same tools.

One pictureMobility, environment, social, economy — silos dissolve.

Trusted evidenceDefensible to a council, bankable to a developer.

Imagine in publicA projective instrument, not just reporting.

Expo modeParticipation as a destination — kiosks, museums, foyers.

The path — about two weeks

Every engagement follows the same arc. The office keeps three things: a documented mandate, a configured platform it re-runs on its own, and a trained team.

What you bring to get started

Starting is light — three things. We handle configuration, training, facilitation and the technical translation.

You bring

Three things

01 · One real questionA live planning question worth piloting — a street, a square, a policy.

02 · A place or a planAn existing site, or a future scenario you want to test.

03 · The peopleA workshop lead from your team, participants in the room, the public online.

For participants: one simple, repeating loop

No plans to read, no software to learn. One full loop takes about half a day, drawn from a sixteen-format method library.

For organisers: source & transform

Behind the room you stay in control. Pick where images come from — upload, library or map — and how participants reshape them: theme, draw or prompt.

The console · 1/3Source & transform.

Set the themes

Define the directions on the table — each with its own prompt, description and expected impacts. Aim the agents at local rules, or let them imagine boldly.

The console · 2/3Set the themes.

Steer it live

During the session, a live gallery with built-in impact assessment lets you moderate the discussion and watch ideas land in real time.

The console · 3/3The live gallery, with impact assessment.

Own the connective layer

A planning decision pulls on the same actors almost everywhere — yet they rarely sit at one table, and the public is usually consulted last. The opportunity is not another plan. It is the connective layer between the actors. No one owns it by default; whoever convenes it sets the standard for everyone else.

What a shared process establishes

The actors in the same visual conversation before positions harden; a method any private plan can be asked to run; and evidence everyone trusts, because every actor worked the same images and votes.

Linkage · process · mandate

A repeatable standard

One tableNot a sequence — the same visual conversation, at the same time.

A standardPropose, see consequences, deliberate, document — plan after plan.

Shared evidenceDefensible to a council, bankable to a developer.

How we build that capability with you

Every engagement starts with the office itself — its real workflows, constraints and ambitions. We map how the office plans today, listen for ambition before features, co-design the surfaces, then pilot on one real question. The office stays the author; we build the instrument.

The question we ask planners: what does it mean, now, to be a planner? The compliance-checking goes to the machine; the human work gets the room it never had. Framework developed with

Where we usually start: innovation agencies

Our entry point is usually the innovation unit — the team with the mandate to experiment before the line organisation commits.