"The organizations participating in Svea demonstrate the bold leadership that Sweden needs."

Within the project, 2,000 people have already been trained in the effective and responsible use of generative AI, and 1,500 public sector employees use the chatbot prototype Svea every week. 400 people have collectively spent 3,000 hours creating 200,000 data points to ensure that the chatbot becomes better at answering questions posed by public sector employees.

Huddinge municipality joined before the second stage and was among the first to decide to continue contributing to the work next year.

"When stage 3 opened, it was a given for us to continue being part of this exciting journey. The project provides us with invaluable knowledge, networking opportunities, and above all, a significant increase in our AI maturity, both at an organizational and individual level," says Linda Burman, innovation leader at Huddinge municipality, and continues:

"I have personally gained a stable foundation, which allows me to take an active role in supporting and guiding other AI initiatives in Huddinge municipality. It is largely thanks to the Svea project that I can now contribute to driving AI development forward in our municipality, and I am convinced that the project will continue to pave the way for many important AI issues that other initiatives in the municipality will benefit from."

Linda Burman also highlights the national perspective and the value of large-scale collaboration:

"Being part of a national initiative aimed at creating a digital AI assistant for the public sector is something we are very proud of. We see it as a matter of course to contribute to a resource that can benefit the public sector," she says.

National AI platforms a critical societal infrastructure

Martin Svensson, managing director at AI Sweden, views national platforms for the public sector's use of generative AI as a long-term critical infrastructure, as necessary and obvious as roads and electricity supply.

Jonatan Permert notes that the development work of the prototype shows that Sweden needs a combination of different types of language models:

"We need to investigate various possibilities for how Sweden can ensure access to and further development of fundamental models for the public sector's needs. Fine-tuning open models is a natural way forward, which makes relationships with major international actors important for Sweden's future. But there is also the development of certain models that we need to do from scratch," he says.

Three challenges for the public sector's use of generative AI

Jonatan Permert points to three central challenges for the public sector's use of generative AI that the initiative is designed to overcome:

1. Competence, a lack of ability to create value with generative AI. In several steps, from the use of tools to the development of the technology.
2. Resource scarcity in several dimensions: AI experts, computing power, and data.
3. Legal aspects, where it is difficult to navigate the intersection of AI and privacy – especially when the technology is to be used with sensitive information.

Martin Svensson emphasizes the importance of collaboration as a solution to these challenges:

"The organizations participating in Svea demonstrate the bold leadership that all of Sweden needs. They are not just investing in a technical solution; they are investing in new shared working methods and building the competence that will define the future of welfare. Together, they show beyond all doubt the strength of collaboration and why it is an absolute necessity for success with AI in the public sector. It is by sharing data, costs, and competence that we build something greater than the sum of its parts," he says.