Written by Craig Grabina
While nearly every business leader claims their organization is AI-ready, more than half admit that data management is not working as well as they should, revealing a critical strategic gap.
Most organizations are investing heavily in AI, with UK businesses currently putting over 20% of their technology budgets into AI. However, Semarchy’s 2026 State of Data Management study reveals major contradictions. While 99% of leaders claim to be AI-ready, 56% admit their data infrastructure is not fit for purpose. This gap in awareness and readiness is why many AI investments fail to meet expectations.
Although most companies claim to be AI-ready, more than half lack the data infrastructure to support AI, revealing a widening gap between ambition, investment, and operational reality.
At the heart of nearly every boardroom conversation about AI is a quiet contradiction. When you ask senior leaders if their organizations are ready for AI, nearly all (99%) say they are. But when you ask these same leaders about the current state of their data management, 56% admit it’s not up to snuff.
Both answers cannot be true at the same time. Yet they are often given by the same person, in the same week, and sometimes at the same meeting.
This gap, derived from Semarchy’s 2026 State of Data Management study, is the most important issue facing AI strategies today. It’s not a difference in technology or a difference in budget. This is the gap between awareness and readiness, which is why many AI investments currently fail to translate confidence into commercial outcomes.
The numbers around it are equally revealing. The research shows that 97% of organizations are actively investing in AI, with UK businesses currently dedicating more than 20% of their technology budgets to AI, but only 5% have appointed a dedicated AI lead. The remaining companies rely on the CTO, CIO, and in some cases the CEO, who are already in charge, to develop strategy on top of everything else they are responsible for. Almost one-third of organizations do not measure data quality before inputting information into their AI systems. And only half document the lineage and explainability of their data in a meaningful way.
This is what the AI readiness really is beneath the headlines: a foundation on which we are spending a lot of money but lacking expert leadership and stress-testing what is being built on top of it.
Cost of building on unstable ground
It’s tempting to treat this as a technical problem. It’s a problem that data teams can solve while the rest of the business does the more exciting work of deploying models. But that framework is the problem. The impact of a poor data foundation is strategic, reputational, and financial, and tends to surface at the moments when an organization can least afford it.
Consider what happens when an AI system trained on fragmented data starts making recommendations at scale. Duplicate customer records become a personalization engine that sends three conflicting messages to the same person. Inconsistent suppliers create a sourcing model that quietly distorts overall portfolio exposure.
Rather than cleaning up cluttered data, AI augments it. Whatever inconsistencies, duplications, and blind spots that exist in the underlying systems will simply be scaled and incorporated into decisions that businesses don’t even realize they’re making. Damage is rarely visible when fired. It accumulates quietly and surfaces in customer churn that no one can fully explain, analytics that management no longer trusts, and audit findings that arrive without warning.
And the cost of correction is disproportionately high. Cleaning a dataset is one thing. Retraining the model, rerunning the analysis, republishing the report, and explaining to the board why the insights you’ve been working on for six months are unreliable is a whole other category of problems.
AI without traceability is just AI
The reputational aspect is even more vivid. Regulatory frameworks such as the EU AI Act, alongside new UK rules and sector-specific rules, increasingly require explainability and traceability as a condition of operation. An AI decision that cannot be traced back to the source data is, from a regulatory perspective, an AI decision that should not have been made.
Organizations without such a pedigree face not only fines but also restrictions on how and where they deploy AI. Trust is a commodity in fields such as financial services, healthcare, and government, and it goes beyond a compliance issue to an existential issue.
None of this is hypothetical. We have seen this pattern of failures play out repeatedly across companies that appeared AI-enabled by any external standard. They had a budget, an executive sponsor, and a pilot. What they didn’t have was a defensible answer as to whether the data underneath could support what was being built on top of it.
What real preparation actually looks like
Organizations defining the next decade of AI have approached data management as a strategic decision rather than a technical cleanup exercise. Treating data quality and governance as managing IT is exactly how the readiness gap arises in the first place.
Authentic preparations tend to have three characteristics. The first is to be clear about your intent before implementing AI: knowing what you are trying to accomplish with AI. It sounds obvious, but it’s surprisingly rare. Many initiatives are reverse engineered from technology. So it’s a model that looks for a use case or a pilot that looks for a business problem. Organizations that gain value are those that start with defined commercial outcomes and work their way back to the required data, governance, and controls.
The second is governance, which is treated as infrastructure rather than checkpoints. In many organizations, governance is the gate through which AI projects reluctantly move into production. By the time it is applied, it has already become a point of friction that slows down delivery, creating the false impression that compliance and speed are mutually exclusive. When lineage, quality scores, access controls, and semantic context move with the data itself, AI initiatives consume managed information from day one. There was nothing ungoverned, so there was nothing left to add.
The third is master data management as a strategic asset. MDM has historically been a back-office discipline, but that framework no longer stands up to reality. With 51% of organizations implementing AI initiatives without an MDM foundation in place and 38% not enforcing data quality standards, the impact is inevitable. The majority of AI investments around the world are being made based on data that cannot determine with certainty whether two customers are the same person. A single, trusted view of critical data (the golden record) is a prerequisite for AI to generate answers that leaders can act with confidence.
What unites these three characteristics is that these decisions are made at the top of the organization and not delegated to lower levels. Companies that get this right stop treating data as a by-product of operations and start treating it as an operational asset on which everything else depends.
Candid questions for senior leaders
When faced with a study like ours, our instinct is to assume that the gap represents someone else’s organization. Very rarely. The 99% who say they are ready and the 56% who agree that data management is not fit for purpose are, by definition, virtually the same group. The disconnect is not happening at the edge of the market. That’s what’s happening in the mainstream.
The honest question for senior leaders is not whether their organization is investing in AI. Almost everyone does. What matters is not whether they feel ready. Almost everyone does. The questions are narrower and more unpleasant. If a regulator or board member asks tomorrow the lineage behind a particular AI-driven decision, will your organization be able to show it? If a model recommends a course of action worth tens of millions of pounds, will executives be able to trace it back to the data they have staked their reputation on? If the answer requires hesitation, all is not as it seems.
The opportunities in AI are real, but the cost of standing still is high. But the competitive landscape of the future will not be defined by who built the strongest foundations that can make that investment worthwhile – who treats data governance, quality and master data management as strategic priorities rather than as technical afterthoughts.
Organizations that recognize the difference between false confidence and true preparedness will set the conditions for subsequent markets. The remaining companies will spend years explaining why their AI investments didn’t deliver as promised, and discovering that the answer was in their data all along.
Craig Grabina 
