How we helped the Bydgoszcz City Hall cut manual reporting and connect city data

Cities face many of the same data challenges as large corporations. Different departments and municipal units rely on different IT systems, formats and ways of organising information. That works until the city needs to see the bigger picture.

City-wide reporting, planning and decision-making require those separate sources to work together. In Bydgoszcz, making that happen meant collecting, checking and harmonising data from across the administration – a process that could take months of work from several full-time employees. The result was a significant operational burden and a long wait for a consistent view of the city.

The question was how to bring this information together without months of manual work. 

Through the Technopark Kielce Startup Booster Poland programme, NeverBlink introduced an AI-assisted integration layer that gives data from different sources a shared semantic structure. The immediate goal was simple: reduce the manual work involved in consolidation while making city data easier to compare, analyse and reuse.

The application developed by NeverBlink for the Bydgoszcz City Hall lets the user browse and analyse municipal data.

Getting data out of a file is not the hard part

One of the things we realised quickly in Bydgoszcz was that extracting data was not the hardest part. The real challenge was recognising its context well enough to connect it with everything else. 

Municipal information comes in many forms. Some sources are already structured, such as statistical databases, spreadsheets and tabular files. Others are reports and documents written primarily for people rather than machines. The subject matter is just as varied: the model developed for Bydgoszcz spans 17 functional areas, including finance, public safety, demographics, education, culture and environmental management, which sums up to over 370 different data cubes.

At that scale, differences in file format are only part of the problem. The harder question is what a piece of data actually means. What does it describe? Which location and reporting period does it relate to? How should it be classified? And how does it connect to information reported elsewhere in the city? 

For experienced municipal employees, many of those answers are obvious because they have worked with the data and reporting processes for years. For software, they are not. A number in a spreadsheet does not explain by itself what it represents or where it belongs in a city-wide model.

That is why simply putting hundreds of datasets in one place is not enough. Without the context needed to interpret them, the result is simply a larger collection of disconnected information. With that context preserved, however, the city can compare information across domains and build a much richer picture of how it is changing over time.

Giving city data a common language

This is where taxonomies and ontologies become useful. In simple terms, they provide a shared vocabulary for describing what the city's data means and how different concepts relate to one another.

Information from different files and systems can then be mapped to the same concept when it describes the same underlying thing. Instead of interpreting every new dataset from scratch, the city can reuse definitions and relationships that have already been established. 

The result is more than a common file format. It is a reusable representation of how the city understands its own information.

That matters because the underlying structure of municipal data often changes more slowly than the data itself. Reporting periods, figures and source files change continuously, but administrative units, statistical categories and relationships between concepts are relatively stable. Once that structure has been formalised, it can be reused when new data arrives.

Our solution lets the user automatically extract data from files while preserving its structure and meaning.

Keeping that semantic layer separate from the applications built on top of it also makes the model easier to maintain it. In traditional IT systems, organisational logic is often embedded in database schemas, application code or individual reporting workflows, which can make even small changes expensive. A separate model can be adjusted as definitions, categories or relationships evolve, without requiring every downstream application to be redesigned.

Removing bottlenecks from integration to analysis 

A semantic model only helps if it can be maintained and applied efficiently. If every taxonomy, relationship and mapping still has to be created, updated and applied by hand, the semantic layer can become a bottleneck in its own right. 

That meant making the model itself practical to work with at scale. In the Bydgoszcz project, we developed tooling that makes semantic structures easier to create, adapt, apply and reuse. Structured sources can be integrated directly, while AI-assisted extraction can bring information from unstructured documents into the same shared model.

Once that foundation is in place, the payoff goes far beyond integration. The relationships between datasets no longer have to be reconstructed every time someone wants to run a new analysis. They are already captured in the semantic layer, making it much easier to combine information that previously lived in separate systems. In Bydgoszcz, for example, spatial and demographic data can be connected to analyse measures such as population density across different parts of the city.

That removes friction from everything that comes next. Analysis and visualisation can start from a shared understanding of the data instead of from another round of manual preparation. More advanced reasoning can build on the same foundation. And over time, BI tools, AI systems and agents can work with data that already carries much of the context they need, rather than having to interpret hundreds of unrelated spreadsheets and documents from scratch.

Our solution lets users explore the data and create interactive charts.

What we are taking away from Bydgoszcz

The project demonstrated that this approach can work at a meaningful scale across many different parts of a city administration. But perhaps the more important lesson is what doesn't have to happen.

Improving city-wide reporting does not require every department to abandon its existing systems and move onto a single platform. A shared semantic layer can connect information across those systems while allowing individual departments to continue working with the tools that suit them.

For Bydgoszcz, the practical goal remains simple: spend less time collecting and preparing data, and more time using it. The project has benefited greatly from the openness and collaboration of the municipal team, and we hope the resulting foundation will continue to support better analysis and new ways of working with city data over time.

Our work on this project is financed from EU Funds through the Technopark Kielce Startup Booster Poland programme organised by the Kielce Technology Park.

Total project cost: PLN 403,054.73

Contribution from the European Funds: PLN 399,348.94.