20VC: Nikesh Arora on the Frontier Model Problem: Breadth vs Depth | The Future of Token Costs | Memory Becoming the Moat | Where Value Accrues: Infra, Models, or Apps? | Why Enterprise AI is Not Ready & Systems of Record vs Systems of Intelligence

Enterprise AI adoption requires a fundamental shift from marginally improving existing workflows to reimagining them entirely through agentic systems. While consumer AI prioritizes breadth and tolerates higher false-positive rates, enterprise applications demand deep context and zero-tolerance for errors, necessitating proprietary data and specialized training. Cybersecurity serves as a critical accelerant in this landscape, where AI tools simultaneously weaponize bad actors and force organizations to consolidate fragmented security stacks into unified platforms. As compute scarcity drives current token pricing, long-term efficiency gains will likely reduce costs, enabling more pervasive AI integration. Leaders must navigate this transition by fostering AI-savvy talent and establishing clear, opinionated AI applications that automate repetitive tasks, ultimately reducing reliance on traditional, process-heavy administrative functions while increasing the demand for technical expertise to manage complex, autonomous infrastructures.