The Enterprise AI Data Crisis: Why Your Data Stack Matters More Than Your Model

Summary

While enterprise leaders race to adopt AI, a critical bottleneck is emerging: fragmented, ungoverned data infrastructure. According to Bavesh Patel, senior vice president of Databricks, the quality and effectiveness of enterprise AI depends directly on an organization's data governance and consolidation. In partnership with Infosys Topaz, MIT Technology Review's Business Lab explores how enterprises are struggling to move from disconnected legacy systems and siloed platforms toward unified, open data architectures capable of supporting meaningful AI at scale.

The challenge extends beyond mere technical infrastructure. Companies need to consolidate structured and unstructured data into open formats, enforce rigorous access controls, and make information accessible across business functions. Without this foundation, organizations risk deploying what Patel calls "terrible AI"—outputs that lack context, trustworthiness, and business value. Rajan Padmanabhan, unit technology officer at Infosys, emphasizes that leading enterprises are tying AI deployment directly to business metrics and governance frameworks, moving away from isolated innovation projects toward measurable outcomes.

As AI agents evolve from copilots to autonomous operators managing workflows and transactions, the data infrastructure challenge becomes even more pressing. Organizations that establish the right data foundation now—combining proper governance with accessibility and context preservation—will be positioned to unlock efficiencies, automate complex workflows, and launch new business lines. The competitive differentiator for most organizations, according to Patel, will increasingly be their ability to consolidate their own data with third-party datasets and deploy them effectively at scale.

• As AI agents evolve toward autonomous operations, the demand for high-quality, accessible, governed data will only increase

Editorial Opinion

This is a critical wake-up call for enterprises still chasing AI headlines without fixing their underlying data plumbing. The industry has spent two years obsessing over model capability and ease of access, but the real bottleneck—and the real opportunity—lies in data infrastructure. Organizations that recognize this shift and invest in consolidation, governance, and accessibility now will pull decisively ahead of competitors still treating data infrastructure as a technical checkbox rather than a strategic advantage.