By Craig Gravina
Nearly all business leaders claim their organisations are AI-ready, yet over half admit their data management isn’t up to the job – exposing a critical strategic gap.
Most organisations are pouring significant investment into AI, with UK businesses now committing over 20% of technology budgets to it. Yet Semarchy’s 2026 State of Data Management research reveals a big contradiction: while 99% of leaders claim AI readiness, 56% admit their data foundations aren’t fit for purpose. This gap between perception and preparedness is why so many AI investments are failing to deliver on their promise.
Most firms claim AI readiness, yet over half lack the data foundations to support it, exposing a widening gap between ambition, investment, and operational reality.
There’s a quiet contradiction sitting at the heart of almost every boardroom conversation about AI. Ask senior leaders whether their organisation is ready for AI, and nearly all of them (99%) will tell you it is. Ask the same leaders about the state of their data management though, and 56% will admit it isn’t up to the job.
Both answers can’t be true at the same time – and yet they are being given, often by the same people, in the same week, sometimes in the same meeting.
That gap, drawn from Semarchy’s 2026 State of Data Management research, is the single most important issue facing AI strategy today. It’s not a technology gap, nor is it a budget gap. It’s a gap between perception and preparedness, and it’s the reason so many AI investments are currently failing to translate confidence into commercial outcomes.
The numbers around it are equally revealing. The survey found 97% of organisations are actively investing in AI, and UK businesses are now committing more than 20% of their technology budgets to it, yet only 5% have appointed a dedicated Head of AI. The remainder are leaning on already-stretched CTOs, CIOs, and, in some cases, CEOs to set strategy on top of everything else they’re responsible for. Almost a third of organisations aren’t measuring data quality before feeding information into AI systems at all. And barely half have documented data lineage or explainability in any meaningful way.
This is what AI readiness actually looks like beneath the headline confidence: foundations that haven’t been stress-tested for what’s being built on top of them alongside significant spend but scarce specialist leadership.
The cost of building on unstable ground
It’s tempting to treat this as a technical concern – something the data team can resolve while the rest of the business gets on with the more exciting work of deploying models. But that framing is precisely the problem. The consequences of poor data foundations are strategic, reputational and financial, and they tend to surface at exactly the moment an organisation can least afford them.
Consider what happens when an AI system trained on fragmented data starts making recommendations at scale. A duplicate customer record becomes a personalisation engine, sending three contradictory messages to the same person. An inconsistent supplier becomes a procurement model, quietly distorting exposure across the portfolio.
AI doesn’t clean up messy data, it expands on it. Whatever inconsistencies, duplications and blind spots existed in the underlying systems are only scaled and embedded into decisions the business may not even realise it’s making. The damage is rarely visible at launch. It accumulates quietly, surfacing in customer churn that nobody can quite explain, in analytics that leadership has stopped trusting, in audit findings that arrive without warning.
And the cost of correction is disproportionate. Cleaning a dataset is one thing. Retraining models, re-running analyses, re-issuing reports, and explaining to a board why the insights they have been acting on for six months were unreliable is a different category of problem altogether.
AI without traceability is AI without permission
The reputational dimension is sharper still. Regulatory frameworks such as the EU AI Act, alongside emerging UK and sector-specific rules, increasingly demand explainability and traceability as a condition of operating. An AI decision that cannot be traced back to its source data is, in regulatory terms, an AI decision that shouldn’t have been made.
Organisations without that lineage don’t simply face fines but they also face restrictions on how and where they can deploy AI at all. In sectors such as financial services, healthcare, and government, where trust is the product, that is more than a compliance issue, it’s an existential one.
None of this is hypothetical. We’ve seen these failure patterns play out repeatedly across enterprises that, by every external measure, looked AI-ready. They had the budget, the executive sponsorship, the pilots. What they didn’t have was a defensible answer to whether the data beneath could support what was being built on top.
What genuine readiness actually looks like
The organisations that will define the next decade of AI have undertaken data management as a strategic decision rather than a technical clean-up exercise. Treating data quality and governance as IT housekeeping is precisely how the readiness gap opens in the first place.
Genuine readiness tends to share three characteristics. The first is clarity of intent or knowing what you’re trying to achieve with AI before implementing it. That sounds obvious, but it’s strikingly rare. Many initiatives are reverse-engineered from the technology – a model in search of a use case, a pilot in search of a business problem. The organisations that get value are those that start with a defined commercial outcome and work backwards to the data, the governance, and the controls required.
The second is governance treated as infrastructure rather than as a checkpoint. In many organisations, governance is the gate AI projects reluctantly pass through on their way to production. By the time it’s applied, it’s already a friction point slowing delivery and creating the false impression that compliance and speed are in opposition. When lineage, quality scores, access controls and semantic context travel with the data itself, AI initiatives consume governed information from day one. There’s nothing to retrofit, because nothing was ever ungoverned.
The third is master data management as a strategic asset. MDM has historically been positioned as a back-office discipline, but that framing no longer survives contact with reality. When 51% of organisations are implementing AI initiatives without MDM foundations in place, and 38% aren’t enforcing data quality standards, the implication is unavoidable: the majority of AI investment globally is being made on top of data that cannot reliably tell you whether two customers are the same person. A single, trusted view of critical data – the golden record – is a precondition for AI producing answers a leader can act on with confidence.
What unites these three characteristics is that they are decisions made at the top of the organisation, not delegated to the bottom. The companies getting this right have stopped treating data as a by-product of operations and started treating it as the operational asset on which everything else depends.
The honest question for senior leaders
The instinct, when faced with research like ours, is to assume the gap describes someone else’s organisation. It rarely does. The 99% who claim readiness and the 56% who admit their data management isn’t fit for purpose are, by definition, substantially the same group of people. The disconnect is not happening at the edges of the market. It’s happening in the mainstream of it.
The honest question for senior leaders is not whether their organisation is investing in AI. Almost everyone is. It’s not whether they feel ready. Almost everyone does. The question is narrower and more uncomfortable: if a regulator or board member asked for the lineage behind a specific AI-driven decision tomorrow, could the organisation produce it? If a model recommended a course of action worth tens of millions of pounds, could the leadership team trace it back to data they would stake their reputation on? If the answer requires hesitation, then all is not what it seems.
The opportunity in AI is real, and the cost of standing still is high. But the competitive landscape ahead will not be defined by who built the strongest foundations capable of making that investment worthwhile – who treated data governance, quality and master data management as the strategic priorities they are, rather than as technical work to be done later.
The organisations that recognise the difference between false confidence and real readiness are the ones that will set the terms of the market that follows. The rest will spend the next several years explaining why their AI investment did not deliver what it promised – and expensively, discovering that the answer was in the data all along.
Craig Gravina
