Umeå universitets logga

The last decade has witnessed significant advances in artificial intelligence (AI), a process driven by new and more powerful algorithms, together with increases in both processing power and amounts of available data. As a result, many tools for adopting AI have become commodities for a broader mass of users. In this paper we focus on an emerging form of AI tools: AI service platforms. The objectives of this paper are to enhance understanding of these platforms’ characteristics and, by providing a typology of AI service platforms, knowledge of their dynamics and evolution. Empirically it is based on a narrative review of literature on AI service platforms considering three dimensions: AI definitions, scope, and novelty. A major outcome is a typology of four AI service platforms, designated analytics, learning, conversational, and distributed platforms. In the presented analysis we discuss the current status and potential future development of these types. The findings increase conceptual clarity of AI service platforms and identifies patterns in their development over time. The latter include ongoing platform convergence leading to the types being increasingly intertwined, and a trend towards what we call ‘lightweight AI’, characterized by easy-to-use, low-code platforms.