Phase 3 - A Shared digital assistant for the public sector
| Reference number | |
| Coordinator | Lindholmen Science Park AB - AI Sweden |
| Funding from Vinnova | SEK 3 000 000 |
| Project duration | December 2025 - October 2026 |
| Status | Ongoing |
Purpose and goal
The aim is to create a national collaboration model for how the public sector can create joint AI services by sharing resources, data and expertise. The goal is to further develop the Svea prototype, strengthen the AI maturity of participating organizations, collect training data and use this to train joint AI models for the public sector, and develop a model for long-term funding, operation and development of a joint digital assistant for the public sector.
Expected effects and result
The project focuses on strengthening the AI maturity of the participants; further developing the Svea prototype for increased stability; developing an agent framework; implementing an upgraded RAG system; broadening national annotation and joint model training; conducting testbed work for sensitive information; and establishing a long-term model for governance, infrastructure, funding and open development.
Planned approach and implementation
The project is implemented in four integrated parts. The change work supports organizations with training and methods for working with generative AI in practice. The software development takes place iteratively, where the prototype Svea is improved based on user needs. Through joint annotation, the participants contribute data that forms the basis for AI model training. In parallel, models for long-term operation and funding are investigated to create a sustainable national solution.