Why AI needs solid data foundations deliver business value

In this short Point of View Paweł sets the scene by raising the question - How can solid data foundations enable GenAI to drive business value?

The problem statement

Many organizations invest in AI without the data strategy, governance, and operating model required to turn pilots into performance. Without trusted data, clear accountability, and responsible use, AI produces inconsistent outputs, increases risk, and struggles to demonstrate ROI. The result is stalled initiatives, fragmented efforts, and growing skepticism from business stakeholders.

The author’s perspective and expertise

Observations and learnings from recent projects

Our observations from recent projects prove that a clear data strategy and effective governance provide the necessary discipline AI needs to perform reliably and at scale. At the core, these such discipline ensures:

• Trusted, high-quality data that is standardized and accessible

• Clear accountability for data ownership, and stewardship 

• Rules for responsible, compliant, and explainable application of AI

However, such discipline by itself does not guarantee impact. Organizations should also strive to create a culture of structured experimentation since Large Language Models, in particular, can produce varying outputs in terms of accuracy, and objectivity. This variability makes continuous validation, refinement, and human oversight essential. Teams that work in short, evidence-driven cycles can mitigate hidden risks and improve performance over time. So how can we put this into practice? Start by:

• Aligning AI to strategy: identify outcomes that matter—growth, cost, efficiency, risk—and select use cases that directly move those metrics.

• Mapping data dependencies: for each use case, define critical datasets, quality thresholds, and controls. 

• Establishing agile governance: create privacy, and security policies, and implement them as guardrails across key data assets and products.

• Building the experimentation muscle: stand up sandboxes, design evaluation protocols, and adopt human-in-the-loop processes to steward data quality.

• Measuring and refining: baseline performance, track costs, and continuously improve models, and data quality to maximise ROI.

When these practical elements are in place, organizations can enable the application of AI to be scaled beyond the initial proof of concept and remain aligned to the wider business strategy. 

Concluding point(s)

To summarise, data strategy and governance should not be seen as a way of restricting the application of AI but as an enabler to drive greater value across the organization.

Another perspective  

This article is part of an extended point of view on how solid data foundations enable GenAI to drive business value. Next, Michael Norejko raises the discussion as to why value-led use cases and fit-for-purpose data are non-negotiable, “Why a robust business case still matters for enabling AI.”