Databricks CEO Ali Ghodsi said that artificial intelligence does not have an intelligence problem but rather a context problem [1].
This distinction is critical for the industry because it shifts the focus from building larger models to improving how AI interacts with proprietary data. If the primary hurdle is context, the solution lies in data management rather than just increasing raw computing power.
Speaking during a CNBC Television interview with Jim Cramer, Ghodsi said the perceived limitations of current AI systems [1]. He said that the challenge for AI is not a lack of cognitive ability, but the inability to access the specific, relevant information needed to provide accurate answers [1].
To address these gaps, Databricks is focusing on how the company leverages AI within its own platform [1]. This strategy involves integrating AI more deeply into the data layer to provide the necessary context for the models to function effectively [1].
Integration is already scaling rapidly across the company's ecosystem. Ghodsi said that 80% of databases on the Databricks platform are now being built by AI agents [2]. This shift suggests a transition toward autonomous data infrastructure where AI manages the underlying systems it relies upon.
During the interview, Ghodsi also said the company's recent fundraising efforts [1]. These financial moves are intended to support the continued development of AI-driven product advances and the expansion of the platform's capabilities [1].
“AI doesn't have an intelligence problem. It has a context problem”
Ghodsi's comments highlight a broader industry pivot toward Retrieval-Augmented Generation (RAG) and specialized data pipelines. By framing the issue as one of 'context,' Databricks is positioning its data lakehouse architecture as the essential bridge that allows general-purpose LLMs to become useful for specific enterprise applications.
