Instagram CEO Adam Mosseri announced that the platform has successfully curbed its artificial intelligence spending by discontinuing inefficient initiatives, describing them as projects that unnecessarily consumed processing tokens. Speaking on Lenny Rachitsky podcast, Mosseri emphasized that managing AI resource allocation has become a standard budgetary function, comparable to allocating graphics processing units or personnel budgets. He noted that while Instagram AI consumption will likely increase as employees adopt more complex agentic workflows, the associated costs are expected to decline as major model providers engage in competitive pricing strategies. Despite anticipating lower prices, Mosseri warned that token burn rates could soon match the cost of a senior engineer salary, prompting him to advocate for structured token caps. He suggested these limits should be scaled according to a team track record of generating positive return on investment, dismissing earlier industry trends like token leaderboards that encouraged excessive consumption as fundamentally flawed. This financial recalibration coincides with a broader structural transformation at Instagram. Mosseri revealed that the company has significantly reduced its average team size from approximately thirteen members to compact operational pods of six to seven individuals. The traditional division of labor has been replaced by multidisciplinary groups primarily composed of generalist software engineers and product staff, supported only by targeted specialists when necessary. By reducing coordination overhead and minimizing committee-driven decision-making, these streamlined units are reportedly executing projects with greater speed and agility. The shift reflects an industry-wide realignment as AI tools consolidate engineering functions and blur traditional role boundaries. As technology firms navigate the economic realities of large language model integration, Instagram approach to token governance and organizational design serves as a practical case study in balancing rapid AI adoption with sustainable operational efficiency.