AI won’t lay your foundation

AtkinsRéalis’ José Gilabert explains why data governance is essential to your business.

Artificial Intelligence (AI) can automate your business, predict events and even write reports, but it won’t lay your foundation.

Think of it this way: before architects design a skyscraper, engineers must lay a solid foundation. Without it, even the most innovative design will crumble. The foundation determines the building's stability, height potential, and longevity. Data governance plays the same role in the world of AI: it creates the solid groundwork, ensuring data is clean, structured, ethical and reliable before AI can build upon it effectively.

Without effective data governance, AI is like a skyscraper built on sand. No matter how advanced your AI tools are, without a proper data foundation, the entire structure becomes unreliable, prone to collapse, and ultimately unusable. A business that fails to govern its data properly isn’t just disorganised, it’s vulnerable to serious risk, where AI becomes a source of liability rather than an advantage. Business leaders must take action to embed data governance practices while enabling AI-driven growth.

The nuances of data governance in the AI era

We live in a time where AI is advancing rapidly. However, in the rush to deploy AI, data governance is often forgotten, already leading to some high-profile failures – for example, AI misdiagnosis in healthcare disproportionately affecting minority patients due to poor data diversity. AI is only as good as the data it learns from, which is why businesses must manage their data effectively and data governance can no longer be an afterthought.

Data governance ensures that data is accurate, secure, ethical and well-managed, enabling businesses to make reliable decisions. Similarly, AI governance sets the development, deployment, and ethical use of AI systems. The two are deeply interconnected: while data governance focuses on managing data, AI governance ensures AI systems make responsible use of that data.

But AI adoption comes with unique governance challenges, which need careful mitigation. For example, AI requires large, diverse datasets, increasing storage costs, complexity and security risk. Rather than accepting this as an inevitability, data governance can enable the set-up of optimised systems.

Many AI models rely on third-party data sources, raising concerns about data origin, ownership and compliance. Data governance should ensure the data is the best available, necessary, and has caveats applied where appropriate. An AI system trained on inaccurate or biased data is at risk of making flawed, unreliable, or even harmful decisions. With AI systems potentially inheriting and amplifying biases present in training data, good governance will ensure bias is avoided, or is at least documented, so that nuance can be applied when using resultant insights.

AI can sometimes operate as a ‘black box’, making it difficult to understand how decisions are made. This lack of transparency can lead to regulatory scrutiny, especially in highly-sensitive industries, such as healthcare. Careful data governance can make the black box less opaque, more auditable and therefore more acceptable to regulators, avoiding a wholesale rejection of AI.

The evolving landscape: a race to keep up

Governments worldwide are racing to leverage AI – regulation is under the microscope as it catches up. The UK favours a principles-based framework, allowing industry-led governance, while others like the European Union (e.g., EU AI Act 2025), adopt stricter, risk-based models. 

Within this fast-changing regulatory environment, businesses must transition from traditional, rigid data governance models to agile, risk-aware, and ethically-driven frameworks that can adapt to evolving business and regulatory demands. Paramount in this is recognising that the ‘why’ of data governance goes beyond regulation. Relegating data governance to merely a compliance issue fails to treat it with the strategic lens it deserves. Instead, it should be viewed as a critical ROI influencer, focusing on the problems it solves and aligning it with business uses to deliver measurable value. 

When put in context, the benefits of data governance are undeniable. Studies show that companies with strong data governance improve AI development efficiency by 30-40%, reducing capital expenditure thanks to fewer model errors and reduced rework. Data governance reduces operational expenditure of AI by 15-25%, through eliminating redundant data processes, improving resource allocation and increasing data reliability.

Fines for data violations can be eye-watering, as seen with Facebook’s $5 billion fine for mishandling data. Good data governance will highlight risks in advance, avoiding costly fines and lawsuits by preventing AI bias and ethical breaches and enhancing data security.

Businesses with strong data governance can scale AI solutions faster, innovate with confidence and unlock new revenue streams through gaining investor and regulator trust. Just as a solid foundation allows architects to build towering structures with confidence, well-governed data enables organisations to construct robust AI systems that can scale, perform reliably, and withstand regulatory scrutiny.

Data governance: how to do it right