Friend or Foe? The Influence of Artificial Intelligence on Human Performance in Medical Chart Coding

Abstract

The impact of artificial intelligence (AI) on jobs has generated considerable discussion and debate. Does AI substitute for worker experience or complement the performance of more experienced employees? This debate has significant implications for economies and societies. We developed an AI solution for medical chart coding in a publicly traded company and evaluated its impact on productivity, as conditioned by coders’ experience, in a field setting. We find evidence that AI improves worker productivity overall, but its impact depends on the type of worker experience. Workers with greater task experience, measured by the number of charts they have coded over the years, gain more from AI. However, workers with greater seniority experience, measured by the number of years they have been working for the company, gain less from AI than their junior colleagues. To uncover the mechanism behind this surprising finding, we conducted a survey of senior workers and found that user resistance caused by a lack of trust is the reason for lower productivity gains from AI. This insight is further confirmed in a lab experiment. This study provides new empirical insights into how AI affects knowledge worker productivity, with important implications for wider adoption and use of AI in knowledge-intensive work.