Databricks CEO on AI Pacing, Cyber Risk, and the Enterprise

AI adoption in the enterprise faces significant hurdles, primarily due to a lack of internal organizational context rather than insufficient model intelligence. While public discourse often fixates on existential risks and the need for "pacing," these concerns are largely disconnected from the immediate, tangible threat of AI-driven cybersecurity attacks. Organizations must prioritize building an "ontology"—a digital graph of internal processes, relationships, and data—to enable agents to function effectively. Databricks CEO Ali Ghodsi highlights that while frontier models dominate headlines, practical productivity gains come from integrating these models with proprietary data and implementing automated security controls. Companies are increasingly shifting toward open-source models and cost-efficient architectures to manage token expenses, moving away from the assumption that the most powerful model is always the optimal choice for every task.