CGI supports Jordbruksverket’s most ambitious AI initiative to date

Jordbruksverket (the Swedish Board of Agriculture) is Sweden's administrative authority for agriculture, fisheries and rural development. The agency works to ensure that Swedish food production is sustainable, profitable and contributes to a vibrant countryside with high standards of animal welfare. As a preparedness authority, Jordbruksverket also plays a vital role in Sweden's civil defense.

To simplify and streamline its operations, Jordbruksverket is exploring digitalization, robotics and AI solutions. Among other initiatives, an AI Lab has been established, led by Jordbruksverket’s AI strategist Anna Miller. The lab is responsible for the development and application of AI within this and other agencies and is tasked with testing, developing and experimenting with AI in a way that is safe, transparent and compliant with laws, regulations and ethical guidelines.

AI ambition under pressure

Jordbruksverket has an ambitious strategy for digitalization and AI. When the agency launched its AI Lab in early 2025, it gathered strong technical expertise. However, it soon became clear that expertise in how AI should be managed and governed in a legal, transparent and ethical manner—otherwise known as AI governance—needed to be added.

The agency is subject to a range of regulations, such as the AI Act, the General Data Protection Regulation (GDPR) and the principle of public access to official documents. This places high demands on security, transparency and the handling of sensitive data in areas closely linked to AI. The agency needed a framework to integrate these requirements into its AI projects from the start without hindering innovation. Furthermore, structures and routines were needed to ensure regulatory compliance, clear roles and the distribution of AI responsibilities.

CGI guides transparent and secure AI innovation

Jordbruksverket reached out to its framework agreement suppliers to add AI governance expertise to its AI Lab. The choice fell on CGI and Jonas Forsman, a CGI Director and Consulting Expert in AI. In May 2025, he joined the AI Lab as an advisor focusing on governance, ethics and regulatory compliance, tasked with developing a sustainable framework for AI governance.

Together with the lab, Forsman developed a test framework for transparency, roles and escalation, as well as ethical guidelines and practical processes for how AI projects should be designed and tested. In his role, he also serves as a link between technical developers and lawyers to translate legal requirements into practical solutions and guide the work on new AI projects.

One of the lab's most prominent innovations is the AI assistant Ragnar, an internal chatbot developed, tested and deployed within Jordbruksverket.

• Local infrastructure: Ragnar runs entirely on the agency's own servers, making it possible to work with sensitive data even if the internet is down—a fundamental part of the agency's preparedness.
• Open source: It’s built on open-source code and uses the language model GPT-OSS, providing transparency in how the model was trained and allowing Jordbruksverket to customize it as needed. We are also continuously testing new models as they become available.
• Risk management: To manage risks such as hallucinations and bias, the lab implemented protective mechanisms in the form of system prompts and automated tests that detect and correct inappropriate responses.

Forsman has also been involved in AI-related education and internal competence development, as well as maintaining contact with the agency's network of AI ambassadors who promote knowledge and support throughout the organization.