How will artificial intelligence (AI) affect the organization and dissemination of knowledge in production? In this paper, we study a combination of experimental and staggered deployment of an AI across 3,000 agents and 4 million tech support conversations in a Fortune 500 enterprise software company. Trained on historical tech support transcripts from all agents, the AI augments technical support agents by providing real-time suggested responses and coaching. We find that the AI deployment increases agent productivity in terms of efficiency (average call duration) and quality (issue resolution rate), with the lowest skilled workers seeing the largest relative gains and deployment shrinking the within-team performance gap. Issue resolution rates increase most across firm, geographic, and team boundaries where diffusion of knowledge is likely more costly, benefiting less-skilled and outsourced agents outside the United States. Consistent with a model of knowledge hierarchies, we see that the reduced cost of knowledge acquisition increases the average team size and the average number of distinct topics solved by each agent in the text of the technical support conversations. Together, these results have implications for the impact of AI on the organization of production and the boundary of the firm.