Not All AI Systems Are Created Equal: A Typology of AI Systems' Performance and Affordance

Abstract

This article explores the implications of AI systems for work and organizations by highlighting the qualitative differences in the unique capabilities and material properties of various AI technologies. While previous research has often treated AI as a generic concept, we argue that different AI systems, such as rule-based, supervised learning, unsupervised learning, reinforcement learning, hybrid, and generative AI, play distinct roles in shaping AI assemblages and therefore in driving organizational impacts. We present a typology of AI systems to capture these varying roles. In doing so, we also emphasize the indispensable but varied roles of humans in enacting AI assemblages and the materialization activities required for their implementation. Our argument highlights the importance of recognizing the inherent design characteristics of AI systems, understanding the sociotechnical challenges and opportunities they present, and considering the different types of human expertise necessary for effective integration.