The AI Pioneer Effect: How Peer Effects in AI Adoption Drive Firm-Wide Productivity Gains

Abstract:  Using an employee-level dataset from a major e-commerce technology company, we document strong heterogeneity in agentic AI adoption across engineering teams despite the absence of technical constraints. We find that team-level adoption is driven by the emergence of AI pioneers— individuals who become unusually heavy AI users early on, relative to their colleagues. Exploiting quasi-random variation in the timing of pioneer emergence within a staggered difference-in-differences design, we estimate the causal effect of agentic AI on business value. We find that both experimental output and the number of experiments successfully rolled out to 100% increased following pioneer emergence. These effects are strongest when pioneers are more senior and most pronounced for the output of their junior colleagues. Finally, the transmission of AI capability from pioneer to team is complementary to the pioneer's domain expertise for specialized tasks (e.g., experiment creation), but less correlated with prior skill for more generalized tasks (e.g., engineering issue tickets).