Databricks’ Ali Ghodsi says most companies don’t need smarter AI models
The Databricks CEO argues that enterprises are sitting on a goldmine of untapped productivity, but the bottleneck isn't AI intelligence, it's their own messy data
Ali Ghodsi has a message for every CEO pouring money into the latest frontier AI model: your problem isn’t that the AI isn’t smart enough. It’s that you haven’t told it anything useful about your business.
The Databricks co-founder and CEO made the case during a Bloomberg Tech appearance that the current generation of AI models already exceeds what most enterprises can actually use. The gap between what these models can do and what companies are getting out of them comes down to one thing: context.
The 90% paradox
Ghodsi pointed to a striking disconnect he’s observed while polling audiences across his 2026 speaking engagements. Only about 10% of respondents believe artificial general intelligence has arrived. But roughly 90% say the AI models they use daily are already smarter than most of their colleagues.
“We don’t need AI to get smarter,” Ghodsi stated. “It just is lacking context.”
The “context” he’s referring to is what he calls “enterprise context”: the proprietary data, institutional knowledge, decision-making workflows, and business logic that make a company’s operations unique.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
The data infrastructure argument
Databricks makes its money helping enterprises organize, manage, and activate their data. Its customer roster includes AT&T, Rivian, Adidas, Mercedes-Benz, Unilever, Virgin, and Bayer.
Ghodsi emphasized that companies need to build their own “enterprise context” and “ontology” rather than assuming they can simply buy a turnkey AI solution. Databricks has been building tools designed for exactly this purpose. Its Genie platform, for instance, is positioned as a way to capture the enterprise data, decisions, and workflows that AI systems need to move from impressive demos to actual automation.
Even if AI advancement paused entirely today, most companies would still have years of productivity gains available to them just by getting their existing data in order.
A decade of organizational surgery
Ghodsi’s estimate for achieving full enterprise AI adoption: potentially a decade, even for technology-centric firms. That timeline reflects the reality that plugging AI into a business isn’t a software installation. It’s an organizational reengineering project that touches processes, roles, incentives, and culture.
As of September 2026, Ghodsi has continued to press this message, arguing that enterprises should focus on practical AI challenges rather than existential risks.