Novice Risk Work: How Juniors Coaching Seniors on Emerging Technologies Such as Generative AI Can Lead to Learning Failures

The literature on communities of practice demonstrates that a proven way for senior professionals to upskill themselves in the use of new technologies that undermine existing expertise is to learn from junior professionals. It notes that juniors may be better able than seniors to engage in real-time experimentation close to the work itself, and may be more willing to learn innovative methods that conflict with traditional identities and norms. However, this literature has not explored emerging technologies, which are seen to pose new risks to valued outcomes because of their uncertain and wide-ranging capabilities, exponential rate of change, potential for outperforming humans in a wide variety of skilled and cognitive tasks, and dependence on a vast, varied, and high volume of data and other inputs from a broad ecosystem of actors. It has also not explored obstacles to junior professionals being a source of expertise in the use of new technologies for more senior members in contexts where the juniors themselves are not technical experts, and where technology is so new and rapidly changing that the juniors have had little experience with using it. However, such contexts may be increasingly common. In our study conducted with Boston Consulting Group, a global management consulting firm, we interviewed 78 such junior consultants in July-August 2023 who had recently participated in a field experiment that gave them access to generative AI (GPT-4) for a business problem solving task. Drawing from junior professionals’ in situ reflections soon after the experiment, we argue that such juniors may fail to be a source of expertise in the use of emerging technologies for more senior professionals; instead, they may recommend three kinds of novice AI risk mitigation tactics that: 1) are grounded in a lack of deep understanding of the emerging technology’s capabilities, 2) focus on change to human routines rather than system design, and 3) focus on interventions at the project-level rather than system deployer- or ecosystem-level.