From AI experiments to real societal impact

Many Swedish organisations have made significant progress in exploring AI, but struggle to turn their ideas into practical benefits. In the project Data-driven organisations – best practices for operationalising AI, AI Sweden brings together industry, academia and the public sector to develop shared approaches that enable faster, safer and more sustainable AI adoption. The goal: to shorten the journey from 18 months to just a few weeks – and unlock the potential of Swedish data.

When AI moves out of the lab and starts creating value within organisations, many face the same challenge: how do you move from proof of concept to production?

This is the central question behind Data-driven organisations – best practices for operationalising AI, a project led by AI Sweden in collaboration with around 50 partners.

“We saw that many Swedish organisations were able to build sophisticated prototypes but lacked the methods to put them into production. This project is about creating the conditions for making AI work in practice,” says Fredrik Viksten, Technical Director at the Department of Electrical Engineering at Linköping University and Technical Expert at AI Sweden.

A model that works across society

The project ran from 2024 to 2025 with a clear objective: to reduce the time from initial idea to a fully operational AI service, while strengthening confidence that AI is being used safely and responsibly.

“For us, the question is not whether AI creates value, but how to realise that value as effectively as possible. There is potential to reduce the cost of delivering societal benefits, both in industry and in the public sector,” says Kim Henriksson, Change Manager at AI Sweden.

To achieve this, the project combined research, industry requirements and legal perspectives. Through shared testbeds and open workshops, it established a practical framework for implementing data-driven ways of working and MLOps – the practices used to deploy, operate and maintain AI systems.

AI Sweden deliberately built the project around collaboration between the private and public sectors. Participants were recruited through AI Sweden’s partner network and the project partners’ own contacts with customers, suppliers and other organisations. The project team also approached organisations that could benefit directly from the project or contribute valuable expertise.

“Involving public sector organisations has been crucial. They push us to think bigger and ensure that AI solutions work even under strict regulatory requirements. When something works in the public sector, we know it will work in industry too,” says Fredrik Viksten.

One concrete example is how legal experts, developers and those responsible for day-to-day operations worked together to address issues relating to data sharing, confidentiality and accountability. By tackling these issues collectively, more organisations can adopt AI without risking breaches of legislation or ethical guidelines.

From testing to impact

The project’s findings have been compiled and shared more widely through white papers, video presentations and other knowledge resources. The results are also evident in the participating organisations.

Region Halland, for example, has transformed its approach to MLOps and shared a blueprint for its work. Red Hat has developed and validated openly available MLOps training materials covering both machine learning and generative AI, while the Swedish Transport Administration has gained a basis for decisions on improving the efficiency of its infrastructure. AI Sweden’s testbed also includes several of the architectures developed during the project, which are now available for partners to explore.

“We wanted to build something tangible that others could use straight away. We are now also seeing positive outcomes that we had not anticipated at the outset,” says Kim Henriksson.

One example is Malmö University, which was not a project partner but implemented a plan developed by the participants – providing an opportunity to test whether the knowledge could be applied beyond the project group.

The next step towards a data-driven economy

The project marks a shift in Sweden’s approach to AI – from individual experiments to shared frameworks. In the longer term, it could contribute to a more resilient, sustainable and competitive economy, with AI becoming an integral part of everyday decision-making.

“The real change will come when organisations learn to use their own data securely. That will enable them to build models that truly reflect their operations,” says Fredrik Viksten.

According to the project team, funding from Avancerad Digitalisering was essential to making the work possible. By bringing together and sharing both practical and theoretical knowledge, the ambition has been to make it easier for more organisations – large and small – to benefit from AI.

“Avancerad Digitalisering has been critical to making this project possible. Together with AI Sweden, we can encourage both private and public sector organisations to explore, test and share knowledge. In doing so, we can help level the playing field and strengthen organisations of all sizes, which is hugely important for Sweden,” says Kim Henriksson.

“Being data-driven is a mindset that everyone will need to embrace. Those who learn fastest will be best placed to succeed,” Kim Henriksson concludes.