Helsinki Summer Streets: Co-designing public space with generative AI

The problem

Urban planning decisions shape how people live, impacting where homes are built, how streets are designed and how neighbourhoods evolve.

Yet planning materials are often highly technical. Zoning maps, two-dimensional drawings and regulatory documents can be difficult for non-experts to interpret. For residents without planning expertise, it can be hard to visualise what a proposed development will actually look like in practice.

In 2023, Helsinki was preparing its Summer Streets initiative, which each summer turns several central streets over to pedestrians by limiting car traffic, adding greenery, and bringing in street furniture. The changes would affect the people who use those streets daily, above all local residents and shop owners, so city leaders wanted their input to shape the redesign rather than simply present it to them once it was done.

Gathering that input well is harder than it sounds. Public consultation tends to rely on technical materials, such as zoning maps, scale drawings and written feedback forms, that are difficult to follow without planning expertise. Residents are asked to picture how a written suggestion, such as more greenery or a safer crossing, would look on their street, and to trust that it will become something workable. When revised plans come back weeks later, shaped by constraints the participants never saw, the link between what people asked for and what they get can be lost, which is a common reason participatory planning leaves residents frustrated.

That gap was the practical problem. Without a shared picture of what a proposal would look like, discussion remains abstract, expectations are hard to test against what is realistic, and people without technical confidence tend to drop out of the conversation. Helsinki's aim was to make participation in the Summer Streets redesign something residents and business owners could genuinely understand and contribute to.

The solution

To make participation more effective, the city's planners ran two community workshops in January 2023, each built around the UrbanistAI platform. One brought together a citizen committee of 15 residents; the other, business owners from the streets being redesigned. In both, the platform let the group turn their ideas into street images during the session itself.

UrbanistAI is a cloud-based platform built by Toretei, an AI company, with SPIN Unit, a Finnish-Estonian research lab focused on urban planning. It uses generative AI, software that produces new images from a written description or a rough sketch, and it works on top of real photographs of the actual streets, so participants are shaping a place they recognise rather than an abstract diagram. The platform can also be set up to reflect a city's own planning rules. It can be trained on local policy requirements, such as where street furniture may be placed or how traffic and pedestrian spaces are regulated, so the designs it helps produce stay closer to what local rules allow.

Within a session, a participant could describe a change in words and watch it rendered as a street scene, or sketch a rough shape that the tool turned into a recognisable object such as a bench, flower box or tree. Either way, the result appeared on a photograph of the street in question, so a suggestion like more greenery, a wider pavement or some outdoor seating could be seen on a familiar Helsinki street within minutes.

Participants worked in small groups led by a moderator across two-hour sessions, with city planners present throughout. Each group produced several alternative visions for set sections of street, compared them side by side, and then voted to shortlist the ones they most wanted to take forward. Because everyone was looking at the same image, the discussion could move from general preferences to concrete trade-offs, such as how much room to give pedestrians and vehicles, or where greenery should go.

The planners then used the shortlisted images to open a practical conversation about feasibility, cost and regulation, explaining where an idea would not work and how a promising one might progress. The tool did not make any decisions. What it gave residents, business owners and planners was a shared, concrete starting point.

The impact

1. The workshop designs shaped the final street
The ideas and feedback from the two sessions were carried into the design of the 2023 Summer Streets and built into the result. Both the European Commission and UrbanistAI record this as the concrete outcome, and the Commission adds that the shared understanding of the visualisations created proved valuable in shaping the final design. This is what makes the exercise co-design: residents' and business owners' input was visible in the finished street.

2. What the approach is expected to deliver
The European Commission sets out several benefits it expects this kind of tool to bring, framed as potential rather than measured. Used well, the approach is intended to widen participation, so residents, businesses and other groups can shape public spaces directly, and the outcome fits more people's needs. The shared visual language is meant to clarify ideas, support more concrete discussion, and help a group reach consensus. When the platform is trained on local planning rules (an optional step), it is designed to flag legal or technical problems earlier in the process. Giving citizens hands-on experience with the technology is expected to demystify AI and build trust in the way planning decisions are made.

Lessons learned

• How a proposal is presented shapes the quality of the input. The change that mattered most was moving the discussion from words to images. Once an idea appeared in a photograph of the real street, people could respond to something specific, including those with no planning background. For any government running public engagement, this points to a broader lesson: the format of the engagement has a significant effect on how useful the input proves to be.
• The tool requires human judgement. The platform produced the visuals, but it did not run the workshops or judge what was possible. Moderators guided the groups, and city planners weighed each idea against feasibility, cost and regulation. A generative AI tool can make a rough idea look finished, so without that professional judgement, realistic images could easily be taken for workable plans. The value came from pairing quick visuals with human expertise about what would actually work.
• Working from the real place kept ideas grounded. Because the images were built on photographs of the actual streets, participants designed within the real constraints of the space rather than for an idealised version of it. That kept proposals closer to what could be built and made the discussion more practical. Tools that start from the real context give participants a more honest sense of what is possible.
• A single, well-defined initiative was a manageable place to start. Helsinki applied the tool to one programme, Summer Streets, with a clear and contained objective. That made the pilot easier to run and to learn from than a broad rollout across all planning would have been. For others, a bounded first use is a low-risk way to test whether the approach fits before committing to it more widely.

Helsinki's experience suggests that generative AI can aid participatory design when it is carefully integrated, honestly framed, and supported by professional judgement.

This case study was written with assistance from artificial intelligence.