Digital technologies, such as AI, are described as “disruptive” (Hampel 2019. Säljö 1999) in learning settings across a range of workplaces, since they trigger innovation and (re)formation of learning practices that target change and improvement. The aim of this study is to explore how strategies for implementation of AI shape learning content and associated social practices in public agencies. In order to gain insight into how public agencies that are vital for inclusive citizenship learn about digitalization and AI, and which knowledges and skills they identify as crucial for their future development, the study focuses on informal learning triggered by strategic initiatives. Informal learning is here understood as the pursuit of insights, knowledges, and/or skills that occurs without the presence of externally imposed curricular criteria (Livingstone, 2006: 206). In this case, the vast field of digitalization and AI form the learning content toward which the participating organizations are orienting their learning needs.

The theoretical point of departure is anchored in the field of lifelong learning, and in the established view of knowledge and experiences as arising in interaction with others and with the environment. These are ideas that were expressed early on by thinkers such as philosopher, psychologist and educator, John Dewey, who has become known for an experience-based conception of learning, what has later come to be called "learning by doing". Talking about learning, rather than education, leads to a shift in focus from formal education to how an individual transforms experiences into knowledge, skills, attitudes and values at all ages, throughout life (Jarvis 2014). In the prolongation of Jarvis’ (2014) reasoning, learning involves experiential dimensions of the enacted social practices. The framing of social practices as transformational in a lifelong learning perspective, regarding knowledges and skills as well as the own professional identity, serves the analysis’ focus on movement. Analytical tools developed in practice theory (Hager and Beckett 2019) are used to identify successive stages in the learning, where the knowledge content progressively is formulated in interaction with participants and researchers.

The empirical material was gathered during a series of three workshops where 11 employees from five organizations analyzed and discussed the situation at their workplace as part of their learning and implementation process. The methodological design is based on an interactive research approach (Ellström, et al., 2020) that emphasizes interactional dimensions between the research sphere and the professional sphere. The study is carried out with full awareness by all parties of two ongoing parallel processes with sometimes overlapping purposes: Emergent learning practices at work and the explorative research process. Based on the empirical material, we noticed that 1) learning needs to be framed by a specific organizational context, 2) that employees appeared to be learning by doing as they were entering unknown territory, and 3) that learning took place within the existing frames of references. To challenge these frameworks to address the identified gaps in the emergent learning practices, we discuss them in relation to the model for responsible AI in practice developed by Dignum (2019).