AI has made the consequences of poor data visible, immediate, and costly.
A decade of promises about digital transformation has collided with an uncomfortable truth: the data that powers artificial intelligence was never truly ready. Across industries, enterprises are discovering that the messy, inconsistent records tolerated by human analysts become catastrophic liabilities when fed into autonomous systems. This is not a technology failure — it is a reckoning with neglected foundations, and it is forcing data governance out of the back office and into the boardroom.
- More than 85% of companies admit that poor data quality is the single greatest barrier preventing AI from scaling beyond small experiments — despite 80% having increased their budgets.
- AI systems, unlike human analysts, cannot compensate for duplicate records, missing fields, or conflicting departmental definitions — errors that were once minor annoyances now cause real-time failures and costly decisions.
- The crisis has elevated once-obscure concepts — metadata catalogs, data traceability, shared semantics — to the center of corporate strategy, reframing data governance as a business imperative rather than a technical chore.
- Companies are turning to AI itself as the remedy, with over 90% of data leaders deploying it for classification, documentation, and quality management — a feedback loop where the problem and the solution are the same technology.
- Only one in four organizations has moved beyond pilot phases, with most enterprises placing measured bets on efficiency gains — time saved, errors reduced, costs cut — rather than sweeping transformation.
Companies have spent heavily on artificial intelligence over the past year, only to encounter an uncomfortable obstacle: their data was never ready for it. More than 80 percent of enterprises increased their data and AI budgets, and 86 percent plan to spend even more — yet 85 percent acknowledge that data quality is the primary barrier to scaling these systems beyond controlled experiments.
The underlying problem is not new, but AI has made it impossible to ignore. Human analysts could work around inconsistent records, duplicate entries, and metrics that meant different things to different departments. Autonomous systems cannot. When flawed data feeds a real-time decision-making agent, the system does not adapt — it fails, or worse, acts on corrupted information. Enrique Manso of EY Spain describes the shift plainly: AI has made the consequences of poor data management visible, immediate, and expensive.
This reckoning has repositioned data governance at the heart of corporate strategy. Traceability, metadata catalogs, and shared definitions — once the quiet concerns of technical teams — are now business priorities. For nearly half of the companies surveyed, data and AI have become the primary engine of growth, not a supporting function.
Paradoxically, organizations are deploying AI to repair the very foundations it exposed as broken. More than 90 percent of data leaders are already using AI for governance tasks: extracting information from unstructured documents, auto-generating metadata, and redesigning architectures to support automation. Some are experimenting with self-governing systems that detect anomalies in real time.
Maturity, however, remains uneven. Only about a quarter of organizations have genuinely integrated AI into their operational data management. The rest are still running pilots, testing what delivers real value and what stalls in the proof-of-concept stage. Success is now measured not in transformation narratives but in concrete returns — efficiency gains, error reduction, time saved. The bottleneck, it turns out, was never the algorithm. It was always the unglamorous work of getting the data right.
Companies have spent the last year pouring money into artificial intelligence, but they've discovered something uncomfortable: they don't actually have the data infrastructure to make it work. More than 80 percent of enterprises increased their budgets for data and AI over the past twelve months. Eighty-six percent plan to keep spending more. Yet 85 percent of them admit that data quality is the single biggest obstacle preventing them from scaling these systems beyond small experiments.
The problem is not new. For years, companies have talked about digital transformation and automation. But those conversations assumed a certain kind of data—messy, incomplete, inconsistent data that humans could work around. A person reading a report can see that a number doesn't quite make sense and adjust. A traditional analytics dashboard can tolerate some duplicates, some gaps, some semantic confusion between departments. AI systems cannot. When you feed bad data into an autonomous agent designed to make decisions in real time, the system doesn't adjust. It fails, or worse, it makes decisions based on corrupted information. Suddenly, the data problems that seemed minor become expensive.
Enrique Manso, who leads artificial intelligence and data strategy at EY Spain, puts it plainly: AI has made the consequences of poor data management visible, immediate, and costly. The technology has forced companies to confront issues they'd been ignoring. Duplicate records, incomplete information, metrics that different departments interpreted differently—these were tolerable annoyances in the age of quarterly reports. They are not tolerable when they feed systems that operate autonomously.
This reckoning has shifted how companies think about data governance. Concepts like data traceability, metadata catalogs, shared semantics, and information quality have moved from the back office to the center of strategy. They are no longer technical problems. They are business problems. For 46 percent of companies surveyed by EY, data and AI have become the primary engine of growth, not a supporting layer. The promise is concrete: better operational efficiency, lower costs, faster time to market for new products. Some of these promises are starting to materialize in financial results.
Interestingly, companies are using AI itself to solve the data quality crisis. More than 90 percent of data leaders surveyed are already deploying AI in some aspect of data governance, classification, documentation, or quality management. The most common applications are automated extraction of information from unstructured documents (36 percent), automatic generation of metadata and catalogs using intelligent agents (34 percent), and redesign of data architectures to support automated systems (27 percent). Some organizations are experimenting with self-governing data systems that detect anomalies in real time and automate routine supervision tasks.
But maturity remains uneven. While experimentation has spread across nearly every sector, only about one quarter of organizations have actually integrated AI into their operational data management. The rest are still in pilot phases, running low-cost proof-of-concept projects and rapid experimental cycles to figure out which initiatives actually deliver value and which ones never leave the testing stage.
The way companies measure success has shifted too. Instead of chasing the grand promises of transformation, they're focusing on measurable returns. Eighty-four percent cite improved operational efficiency as the primary benefit of AI, far ahead of revenue growth or customer experience. In mid-sized companies, 75 percent expect AI to directly reduce costs, and 77 percent link it to customer experience improvements. Data leaders say the initiatives that build the most confidence are those focused on time savings, process automation, error reduction, and productivity gains—concrete, visible, immediate benefits rather than strategic transformation.
The picture that emerges is one of pragmatism replacing hype. Companies are investing heavily, but they're learning that the bottleneck is not computing power or algorithm sophistication. It's the unglamorous work of getting their data in order. They're moving forward through small, measurable experiments, learning what works, and building confidence one efficiency gain at a time.
Notable Quotes
AI is making visible, immediate, and costly the consequences of not managing data well— Enrique Manso, EY Spain AI and Data Strategy Lead
Data and AI have become the primary engine of growth for 46% of companies, not merely a technical layer— EY survey of Chief Data Officers across 40 countries
The Hearth Conversation Another angle on the story
Why did companies suddenly realize their data was broken? They've had data for decades.
Because AI doesn't tolerate what humans do. A person can read a report with a typo or a duplicate and understand the intent. An autonomous system can't. It just processes what it's given.
So the technology exposed a problem that was always there?
Exactly. The problem was always there. It just didn't matter much when humans were the final reader. Now that machines are making decisions, it matters enormously.
Is this slowing down AI adoption?
Not really. Companies are still spending more. But they're being more realistic about what they need to do first. They're fixing the foundation before building the house.
What's the most surprising finding in the data?
That 90 percent of companies are already using AI to manage their data problems. They're not waiting for perfect data. They're using the technology to create the conditions the technology needs.
That sounds circular.
It is, a bit. But it's working. The companies making real progress are the ones treating data governance as a business priority, not a technical checkbox.
What happens to the companies that don't fix this?
They stay in pilot mode. They run experiments that never scale. They spend money without seeing returns. Eventually, they either invest properly in data infrastructure or they give up on AI.