The modern data stack has seen substantial success, generating over $100 billion in market cap by focusing on structured data through a cycle of ETL, transformation, and analysis. However, as businesses increasingly rely on AI, the need to integrate unstructured data—like transcripts, reports, and web searches—has become critical. The semantic layer, which unifies metrics across data types, must now adapt to handle both structured and unstructured data at scale. This evolution is essential since unstructured data powers AI and its applications. Recent acquisitions highlight this shift in focus. Datadog's purchase of Metaplane, Snowflake’s acquisition of Observe for $1 billion, and ClickHouse’s acquisition of Langfuse all emphasize the importance of data observability in AI systems. These moves are not just about expanding capabilities; they reflect a strategic understanding that monitoring how AI is used will inform product development and customer success. The volume of data involved is massive, and managing it effectively will sustain the rapid pace of AI advancements. Looking ahead, the consolidation of the data stack is expected to continue, integrating more pieces of the AI framework, including evaluations and agent orchestration. As businesses navigate this landscape, the focus will be on creating a seamless blend of technologies that enhance the usability and effectiveness of AI systems.