Why Fragmented Data Estates Can Drain Leaders' AI Budgets
Over 75% of senior technology leaders report that less than 20% of their data is ready for reliable AI agent use, with context fragmentation emerging as a critical budget drain.
Context has become the most overworked word in enterprise AI, with few organizations agreeing on its meaning. According to Louis Landry, CTO at Teradata, context refers to what an AI agent must know before acting—what data exists, what it means, how it relates to other data, when it's valid, and where to find it. This context must be built into the data estate rather than assumed to come with the AI model.
Without proper context, AI agents fall back on raw intelligence, searching, guessing, and inferring information that could have been provided directly. This approach costs money through wasted computational tokens and increases the risk of errors. A 2026 Teradata-commissioned survey from Wakefield Research reveals that more than 75% of senior technology and data leaders estimate that 20% or less of their data is 'ready for AI agents to use reliably.'
The same percentage of leaders reported struggling to unify data and knowledge across business functions sufficiently for agents to reason across the enterprise. These gaps don't disappear when agents gain data access; instead, they become operational risks that may surface weeks later, unlike human analysts who might question suspicious data immediately.
Access Is Not Understanding
If you're relying solely on powerful models and prompt engineering to get the most out of your data, you're doing it wrong.
Context fragmentation—data lacking the meaning, lineage, and governance agents need to act reliably—poses a significant challenge. Unlike humans who might question anomalies, AI agents may act on incorrect information without immediate detection. Organizations must ensure their data carries sufficient meaning for reliable utilization, as more search time increases the potential for AI hallucinations.
Related
