Abstract

As deep learning and big data increasingly shape modern artificial intelligence (AI) tools, it is essential to consider the broader impact of integrating AI into workplaces. While AI applications can optimize processes and improve productivity, their long-term effects on workers’ learning curves and overall performance are still underexplored. This paper investigates the intricate relationship between AI-enabled technology and workers’ learning dynamics through a large-scale randomized field experiment conducted on the Instacart platform. In the experiment, gig workers were randomly assigned to either a treatment groupe with access to AI-enabled tools or a control group without such access. Our empirical analysis reveals that while AI-enabled technology enhances immediate performance by improving both productivity (as indicated by shorter picking times) and quality (through reduced refund rates) of item picking tasks, it may also lower workers’ ability to learn from repeated task experience when having access to the AI tool. This suggests that constant reliance on AI tools can limit hands-on experience and reduce opportunities for independent problem-solving, which are essential for long-term skill development. Furthermore, we examine the carryover effect of AI-enabled technology on workers’ learning curves when access to the technology is no longer available. Workers exposed to AI-enabled tools initially show a productivity performance drop when transitioning to environments without these tools. However, they exhibit higher learning capability, which may enable them to close the initial performance gap and accelerate their learning process when AI support is no longer available. These findings suggest that, while AI can initially hinder learning, it may have the potential to promote deeper cognitive development when managed carefully in the workplace. Finally, our heterogeneous analyses reveal a reduced substitution effect between AI and learning with access to AI, along with a more pronounced benefit of AI on learning in the absence of AI access, when workers have greater prior task-specific experience or have been exposed to similar work environments. In addition, AI-enabled technology tends to facilitate learning more effectively for workers in urban areas, revealing a potential rural-urban divide in AI utilization for skill development. Overall, our findings highlight the importance of carefully balancing the use of AI tools to support both immediate performance gains and long-term skill development in the evolving future of work.

Note: Research papers posted on SSRN, including any findings, may differ from the final version chosen for publication in academic journals.  

---

• Keyword(s):
• ai 24
• operations 148
• workforce development 66

View Working Paper