Artificial intelligence (AI) can potentially contribute to solutions to some of the major challenges facing healthcare systems around the world. However, the development of new AI applications does not guarantee that they will be routinely used in clinical practice. It is well established in implementation and innovation research that the introduction of new technology is associated with various forms of obstacles and resistance. This contributes to delayed and variable application and a high risk of losing the potential value they can bring to the healthcare system in terms of quality, safety, accessibility and optimization. In this presentation, we provide an overview of state of the art regarding the implementation of AI applications in healthcare today. We also suggest ways forward, how to develop successful strategies for implementation and work structured with implementation processes.

To be able to develop strategies and methods for the successful implementation of AI applications in healthcare, it is important to review the accumulated knowledge found in the scientific literature as well as understand the different perspectives that exist among stakeholders in healthcare such as patients, healthcare staff, leaders and politicians. Our research group has carried out extensive studies of available scientific literature and interviews with stakeholders within the healthcare system. In parallell, we also carry out several co-production projects around the development and implementation of AI applications together with health care actors and companies behind the applications.

Our results show that  research has mainly been focused on the application of medically oriented applications where the AI component is based on a low action autonomy , i.e. the AI applications do not differ significantly in their intended application or technical design from other healthcare information systems. Research on implementation has been more focused on studying the implementation of each individual AI application, rather than on a general level trying to understand and promote such implementation processes. The implementation challenges highlighted by different stakeholders are mainly associated with questions about ethical and legal problems and new uncertainties in clinical everyday life created by the introduction of AI applications. Additional challenges that are highlighted around the implementation process are to a large extent associated with the internal capacity for strategic change management and the dependence on healthcare's various professions and practices changing as part of the application of AI applications, as well as conditions that healthcare cannot fully control themselves.

The results point to the need to develop implementation strategies among healthcare organizations to address challenges associated with AI-driven development of quality and capacity. Methodologically designed frameworks for implementation are needed to guide and promote practical application based on overall implementation strategies that include cooperation between the various actors in healthcare and business.