How a Finnish City Built an AI Feature With a Local Company: Our Development Partnership With the City of Joensuu

In 2025 the City of Joensuu selected Rebooted Solutions Oy as a development partner to design and build the first proof-of-concept version of an AI-assisted feedback system. The engagement was run as a direct procurement, delivered iteratively along agile principles, with a target of completing the initial version by the end of the year. It is a matter of public record, documented in the City of Joensuu's procurement decisions.

On paper that is one short procurement decision among hundreds a city makes in a year. In practice it is a small example of something more interesting: a Finnish municipality choosing to build with a local AI company rather than buy a finished product from a large vendor. We think the model deserves more attention than it gets — both from other cities and from the companies that could be doing this work.

What a development partnership actually is

In Finnish public-sector language this is a *kehityskumppanuus* — a development partnership. It is different from the two procurement modes most people picture. It is not a competitive tender for a fully specified system, and it is not buying a shelf-ready SaaS product. It is the city and a company developing something together — in this case starting with a proof of concept, built iteratively — where part of the value is the working solution and part of it is the learning the city gains along the way.

Joensuu has been explicit about inviting this. The city has told companies, at events and in its own communications, that if they have an idea, a plan, or a solution that could fit the city's services — and that could grow into a marketable product — they should get in touch with the city's IT department about possible development collaboration. That is exactly the door we walked through: we approached the city's IT department with a proposal, rather than waiting for a tender to appear.

The "marketable product" part is the quiet point that makes this model work for everyone. The city gets a solution built around its real needs. The company gets a reference engagement and, potentially, the seed of a product it can offer to other municipalities. And the region keeps the work — and the resulting capability — local.

Why a municipality would build with a small local company

It is a fair question. A two-year-old company is not the obvious choice next to established vendors. But for a focused AI feature, size cuts both ways.

• Proximity. We are in Joensuu. We can sit in the same room as the people who actually use the system, understand the workflow, and iterate without a layer of account managers in between.
• Speed and focus. A small, senior team building one thing moves faster than a large vendor allocating a fraction of several people to a side project. There is no handoff between the people who scope the work and the people who write the code.
• The right size for an experiment. AI features benefit from being tried, measured, and adjusted before they are scaled. A development partnership is structured for exactly that — you are not committing to a five-year platform contract to find out whether an idea works.
• Regional value. Public money spent with a local company stays in the regional economy and builds local AI capability. For a city that has publicly stated an ambition to be business-friendly, that is a feature, not a footnote.

How we approach public-sector AI work

Building AI for a city is not the same as building it for a startup. The constraints are stricter, the data is more sensitive, and the bar for "good enough to ship" is higher because the users are residents and public servants, not early adopters who tolerate rough edges. A few principles guide how we work here.

Data sensitivity comes first. Public-sector data — and anything touching residents' feedback — has to be handled with clear boundaries about where it goes and what processes it. That shapes architecture decisions from the start, including which parts of a system can use external AI services and which need to stay inside controlled infrastructure.

A human stays in the loop. Language models are useful for sorting, summarizing, and routing feedback. They also make confident mistakes. For anything that affects how a resident's input is handled, the design keeps a person in control of the decisions that matter — the AI does the heavy lifting, not the final judgment.

Production-grade, not demo-grade. It is easy to build an AI demo that impresses in a meeting and falls over the first time real data hits it. We build the unglamorous parts — input validation, error handling, logging that a non-developer can read, and clear alerts when something is not working — because in a public service, a silent failure is worse than no feature at all.

Build to be handed over. A development partnership should leave the city more capable, not more dependent. That means documentation, clean code, and decisions a future maintainer can understand — so the city is never locked into a single supplier just to keep the lights on.

What this signals for public-sector AI in Finland

Most of the conversation about AI in the public sector happens at the level of national strategy and large platform procurements. The Joensuu model points at something more practical and more repeatable: a city identifying a concrete, bounded problem, partnering with a capable local company to build a focused solution, and treating it as an experiment that can grow if it works.

It is not the right model for every project. Core systems and large-scale platforms still belong in formal procurement. But for the long tail of "this specific workflow could be a lot better with some AI" — the feedback systems, the routing, the document handling, the internal tools — a development partnership with a focused local company is often faster, cheaper, and lower-risk than the alternatives. And it builds regional capability instead of exporting the budget.

We would like to see more Finnish municipalities try it, and more local companies step up to do the work. Joensuu has shown that the door is open.

Rebooted Solutions is an AI consultancy and development partner based in Joensuu, North Karelia. We work with municipalities and public-sector organizations on AI features that are built to ship and built to hand over. If your organization has a workflow that could be better with AI, get in touch — we are happy to talk through whether a development partnership is the right fit.