Enterprise AI ambitions are running ahead of the data foundations required to support them. That is the central argument of a new podcast episode produced by MIT Technology Review in partnership with Infosys Topaz, featuring a senior executive from Databricks.

Bavesh Patel, senior vice president for go-to-market at Databricks, says most enterprise data remains locked inside disconnected applications and legacy systems. "The quality of that AI and how effective that AI is, is really dependent on information in your organisation," Patel says. Without clean, accessible, well-governed data, he argues, companies risk what he calls "terrible AI."

The observation is not new, but the framing is pointed. As AI agents move from assistive tools toward autonomous operators managing workflows and transactions, the tolerance for poor data quality narrows considerably.

What this signals

The piece reflects a broader pattern visible across enterprise software: the bottleneck in AI deployment has shifted from model access to data readiness. Most large organisations can now reach capable foundation models via API. The harder problem is feeding those models accurate, contextualised, access-controlled information drawn from years of fragmented internal systems.

Rajan Padmanabhan, unit technology officer for data analytics and AI at Infosys, frames the shift in architectural terms. He describes a move away from "systems of execution" and "systems of engagement" toward what he calls "systems of action," where AI agents do not merely surface information but complete tasks autonomously.

For that to work reliably, enterprises need unified data architectures that handle both structured and unstructured content, preserve real-time context, and enforce rigorous governance. Padmanabhan also points to AI literacy as an underrated gap: business users, he says, are eager to understand what AI actually requires of their organisations, not just what it promises.

The episode is sponsored content, which limits its independence. Neither Databricks nor Infosys disclosed specific customer outcomes, deployment numbers, or benchmark figures during the conversation. The claims are directionally consistent with what data platform vendors have argued for several years.

Databricks, which raised $15.3 billion across its funding history and was valued at $62 billion in its 2024 Series J, has a clear commercial interest in positioning data infrastructure as the critical layer beneath any AI initiative.