From data ambition to data discipline

Why governance, simplification and leadership realism now define AI readiness

For much of the last decade, the corporate data conversation has been dominated by scale: more data, more systems, more analytics, more ambition. What is now emerging, across sectors as diverse as banking, transport, pharmaceuticals and advanced manufacturing, is a quieter but more consequential shift. Senior leaders are no longer asking how to extract more value from data in the abstract. They are asking how to make data governable, intelligible and trustworthy in an environment defined by regulatory pressure, architectural sprawl and accelerating expectations around AI.

This was the central theme of a Food for Thought lunch, hosted by HotTopics and in partnership with Confluent, in Frankfurt. Part of the wider Data Visionaries community, this rich and fascinating debate moderated by Editor Peter Stojanovic allowed for attendees to share insights, challenges, and consider their priorities for 2026.

Complexity as business-as-usual is bad business

The most striking theme from recent cross-industry conversations is not technological novelty, but organisational realism. Almost every large enterprise is living in at least two worlds simultaneously: legacy and cloud, batch and real time, SAP and non-SAP, centralised control and local autonomy. For some, those worlds have been running in parallel for decades. The challenge is not that this complexity exists, but that it has quietly become normalised. Complexity has become background noise, even as it undermines speed, confidence and compliance.

Nowhere is this more evident than in highly regulated sectors. Financial institutions, transport operators, and life sciences firms are all pursuing cloud migration and advanced analytics, yet remain constrained by data sovereignty, supervisory expectations and sector-specific rules. The result is often a patchwork estate: some workloads in the cloud, others immovable; some data governed tightly, other data effectively orphaned. From a regulatory perspective, this is more dangerous than inertia. 

If leaders cannot state with confidence where their critical data resides, how it flows, and who consumes it, then compliance becomes an exercise in hope rather than control.

This is why the current enthusiasm for generative and agentic AI is being tempered by a more sober assessment of readiness.

Executives increasingly recognise that AI systems are not magical layers that sit above organisational reality. They are amplifiers. If the underlying data is inconsistent, poorly defined or weakly governed, AI will simply scale those flaws, often in ways that are harder to detect and explain. Several practitioners noted that the hardest work is not model selection or tooling, but “agreeing a common language for data across the enterprise”, or, what a customer is, what an order is, what constitutes “clean” or “complete”. 

Without that semantic alignment, automation becomes fragile and insight becomes contested.

One leader described how progress only became possible when data quality was made an explicit performance objective across both IT and business teams. Cleaning data stopped being a technical aspiration and became a managerial expectation. The lesson for the C-suite is therefore uncomfortable but clear: data maturity is a leadership issue before it is a technology issue. Incentives, accountability and cultural permission matter as much as platforms.

There is also a growing recognition that simplification does not mean standardisation at all costs. 

Many enterprises are consolidating aggressively, retiring unused applications, reducing on-premise estates and forcing greater transparency around the true cost of systems. Making business units directly accountable for the systems they insist on retaining has proven to be a powerful, if politically sensitive, lever. Simplification, in this sense, is not about elegance. It is about focus.

Explosion of data undercuts risks

At the same time, leaders are grappling with an explosion of operational data. Sensors, connected assets, localisation systems, and customer interaction platforms are generating continuous streams of information, often with wildly different latency and quality characteristics. Some data is valid for seconds, some for years. The strategic question is not how to store all of it, but how to distinguish between signal and noise in real time, and how to ensure that downstream systems receive only what they need, when they need it. This is as much an architectural concern as it is a financial one, given the rising costs of data movement and duplication.

Security and resilience cut across all of these issues. Executives with operational responsibility are increasingly blunt: it’s not a question of if, but when a cybersecurity incident will occur. The emphasis is shifting from perimeter defence to containment, from preventing every breach to ensuring that no single compromise cascades across systems. This again reinforces the importance of well-defined data boundaries, access controls and observability. Poorly understood data flows are not just inefficient; they are a systemic risk .

Finally, geography matters. 

Global organisations must reconcile radically different data ownership regimes across regions, from customer-owned data in Europe to state-owned data in parts of Asia. The most resilient strategies are those that establish a common conceptual and architectural template, while allowing for local legal and operational variation. Uniformity of thinking, not uniformity of deployment, is emerging as the winning model.

Closing thoughts

For the C-suite, the implication is clear. The next phase of data strategy will not be defined by bold experimentation alone, but by disciplined execution. The organisations that win will be those that treat data as infrastructure rather than exhaust, governance as an enabler rather than a brake, and simplification as a continuous leadership commitment. In a world rushing towards AI, the competitive advantage may belong to those who slow down just enough to get the foundations right.

This Food for Thought was made in partnership with Confluent