Large interview study on municipal AI usage

Lärdomar från Kraftsamlingen: Vad krävs för att lyckas med AI? (Lessons learned from Kraftsamlingen: What is required to succeed with AI? is the title of a new report from AI Sweden. It is based on interviews that students at the University of Gothenburg conducted with municipalities that have participated in various ways in Kraftsamlingen (Collaboration for AI in municipalities and civil society).

“Through Kraftsamlingen we have supported very concrete projects, but also worked with strategic and structural challenges from the municipalities, including through the AI Council that we started together with the Swedish Association of Local Authorities and Regions (SKR),” says Rebecka Lönnroth and continues:

“Thanks to the students, we had the opportunity to go in-depth on some of these issues. What works? What is the challenge? What is the homework a manager in a municipality needs to do?”

The compiled answers point, among other things, to the need for a more systematic approach and the great value that can be gained from concrete collaborations, so that one does not merely stop at what the report describes as "collaboration through discussions".

Tomas Lindroth, Senior Lecturer and Head of Department of Informatics at the University of Gothenburg, supervised the students and co-authored the report.

Rebecka Lönnroth elaborates on this reasoning:

“There is some talk about pilot cemeteries. But one must remember that the public sector needs to explore how technology should be used. Problems arise when good pilots are not taken further, when the implementation of the solution has not been considered from the start. Or when someone repeats what others have already tested instead of talking to other municipalities and building on what they have already done. There is much to be gained here from structure and collaboration,” she says and continues:

“What becomes particularly costly and resource-inefficient is if we have many silos out in the organization with projects that are driven by enthusiasts and are not part of something larger. Then managers lose control over what is happening in the operations related to AI. A more systematic approach makes it more resource-efficient within the organization, which can be further scaled up if one joins forces with others and makes joint investments with the resources one has.”

Tomas Lindroth says that the need for exploration fits poorly into the traditional ways in which public activities are governed, and in how procurements are carried out.

“It is different to procure an established technology, such as a medicine robot for elderly care, which is known to create value, and to start working with language models where it is not yet obvious where and how the great gains will arise. But here, the sector itself has a great responsibility to find the answer to that. The challenge is that the procurement processes are designed to minimize risk,” he says.

The report also highlights the need for courage. A manager who understands the technology in relation to their own operations feels greater security and integrity and is therefore better equipped to make decisions that outsiders might perceive as "courageous."

“But being a manager is often lonely. You bear the risk yourself, especially in a politically governed organization, it can be very scary. But through collaborations, you find people to link arms with, and in this way also spread the risk. So my appeal to managers is to find contexts where you can learn together with colleagues.”