Data strategy: the decisive factor in AI leadership.

A striking theme in enterprise technology news this month is the push to break down data silos as a means to unlock AI at scale. Many companies large and small are grappling with fragmented data landscapes – legacy systems and disparate databases that don’t talk to each other ([1]). The consequence of this fragmentation is that data often remains “locked” in functional or departmental silos and in incompatible formats, stripped of the business context that AI needs to derive insight ([2]). This lack of integration hampers AI initiatives: models trained on narrow, isolated datasets struggle to perform in real-world scenarios, and promising pilot projects can’t scale due to slow data integration and compliance concerns ([3]). Simply put, an AI is only as smart as the data it can readily access.

To address this, leading organizations are overhauling their data architectures with an emphasis on openness and unification. One approach gaining significant traction is the "data lakehouse" – a modern data architecture that combines the rigorous data management of warehouses with the flexibility of data lakes. A recent market study by Forrester found that data lakehouses have become the default architecture for modern analytics and AI workloads ([4]). The reason is straightforward: they simplify data architectures, reduce costly duplication, and allow enterprises to tap into both structured and unstructured data in one environment. By eliminating unnecessary data movement, teams can train and deploy AI models directly on comprehensive, up-to-date datasets, improving both speed and accuracy of insights.

Major tech investments in the last fortnight reflect this shift toward unified data foundations. For example, enterprise software giant SAP announced a deal to acquire data platform Dremio, aiming to merge data across its SAP Business Data Cloud and any external sources seamlessly ([5]) ([6]). Dremio is a leading proponent of open data formats (notably Apache Iceberg) and "enterprise lakehouse" technology, which will let SAP customers query and analyze SAP and non-SAP data together without cumbersome extract-transfer-load processes or data duplication ([7]). SAP itself acknowledged that many AI projects fail not due to AI, but because underlying data is fragmented and locked away in proprietary systems ([8]). By moving to an Iceberg-based open table format, SAP is effectively saying that enterprises can no longer afford to have their data in walled gardens. The goal is a single, federated data layer where every analytics and AI tool has consistent, governed access to the full spread of enterprise information ([9]). This kind of unified architecture is poised to become a competitive necessity, as companies that can quickly integrate new data sources and maintain a holistic view of their business information will innovate faster and scale AI with far less friction.

Another key differentiator emerging among AI leaders is the ability to harness data in real time. Many AI-driven services – from personalized customer experiences to predictive maintenance and automated decision-making – require immediate, up-to-the-minute data rather than static batches. Forward-looking enterprises are thus investing in streaming data pipelines and systems that feed their machine learning models with live data flows, so their AI can react to events as they happen. A prime example came with IBM’s recent announcements: at its flagship conference this month, IBM’s CEO stressed that companies winning in AI are not merely deploying more algorithms but “redesigning how their business operates,” supported by a real-time, AI-ready data foundation ([1]) ([2]). To put weight behind this vision, IBM made one of its largest acquisitions since 2019 – a $11 billion deal to purchase Confluent, a leading cloud data streaming platform ([3]). By bringing Confluent’s Kafka-based streaming technology into its watsonx data platform (now dubbed “IBM Confluent”), IBM is enabling enterprise AI systems to access fresh, trusted data from across disparate sources instantly, rather than waiting for overnight updates ([4]).

In parallel, organizations are adopting new types of databases designed for AI’s unique needs. Chief among these are vector databases, which store data (such as text, images, or customer records) as high-dimensional vectors to enable semantic search and retrieval for AI models. The booming interest in generative AI has turned vector stores from a niche tool into a centerpiece of modern AI infrastructure. In fact, Gartner projects that by 2026, more than 70% of all generative AI applications will rely on vector databases to handle unstructured data and provide context for AI queries ([5]). This reflects a shift in enterprise data strategy: companies recognize that to get meaningful answers from AI (for example, a chatbot that can draw on all of a company’s documents or real-time data), they need new data layers that can supply contextual knowledge at machine speed. Cloud and database vendors are racing to integrate vector search capabilities, and early adopters report significant gains in search relevance and faster AI-driven insights. The message is clear – AI-ready infrastructure now means streaming and vector-enabled data systems that ensure no information is out-of-date or out-of-reach when the algorithm needs it.

As artificial intelligence technology becomes more widespread, the competitive playing field is being defined by who owns the best data. In contrast to past years when owning a superior algorithm or model was seen as the key advantage, today’s executives are increasingly treating proprietary data as invaluable intellectual property – the “moat” around their business. We’re seeing real deals and partnerships formed for the primary purpose of securing unique data assets. At a recent industry gathering, experts highlighted how top AI firms are partnering with organizations that have rich, domain-specific datasets, underscoring that durable differentiation lives in the data underneath the model ([1]). For example, the AI company Anthropic has begun co-developing AI solutions using the internal knowledge repositories of global law firm Freshfields, in a multi-year collaboration to build AI-driven legal workflows ([2]). By training on Freshfields’ vast trove of legal documents and expertise, Anthropic can create more specialized legal AI models while Freshfields gains early access to cutting-edge AI – a symbiotic arrangement built on a shared data moat.

Investors and markets are also recognizing the outsized value of unique data. One analysis notes that companies with proprietary, high-quality training datasets are commanding 3–5 times higher valuations than competitors without those data assets ([3]). The rationale: as foundational AI models become commoditized or open-sourced, it’s the exclusive data – customer interactions, domain-specific knowledge, real-time user behavior, or industry-specific troves – that can’t be easily replicated. Organizations that own such data can continually refine AI offerings in ways others cannot, leading to better insights and services. Conversely, firms that rely heavily on third-party or public data may find themselves at a strategic disadvantage. They risk having no unique training material to distinguish their AI, and worse, may inadvertently leak valuable information to the very platforms that could arm their competitors ([4]). This dynamic is encouraging more companies to keep a tight grip on their data exhaust and to invest in data collection, annotation, and stewardship as core competencies. In the AI era, your data isn’t just fueling your models – it is becoming the secret sauce that sets your business apart.

The rise of AI as a business driver is being matched by a rise in regulatory and ethical scrutiny of data. Around the world, governments are introducing and refining laws to ensure AI developments don’t compromise privacy, safety, or transparency ([1]). Europe in particular has been active: negotiators recently agreed on adjustments to the upcoming EU AI Act to give companies slightly more time to comply with requirements for “high-risk” AI, while accelerating transparency obligations ([2]). The message from regulators is that robust data governance is not optional. Companies will need to document data sources, ensure quality and fairness of training data, and maintain strict controls on data usage – or face hefty fines and legal challenges. Indeed, since GDPR came into effect, data privacy penalties in Europe have surged, totaling over €7.1 billion (with €1.2B issued in 2025 alone) ([3]). Data mismanagement is evolving from a hidden technical debt into a direct compliance and business risk.

Leading enterprises are responding by making data sovereignty and governance a top C-suite priority. In many sectors – from finance to healthcare – companies are moving towards “sovereign cloud” solutions that keep sensitive data and AI processing confined within specific regions or jurisdictions to meet local laws and customer expectations ([4]). Just last week, Workday announced a dedicated EU Sovereign Cloud aimed at European clients, ensuring that all their HR and financial data, along with any AI computations on that data, remain strictly in-region ([5]). Such measures go hand-in-hand with building trust: customers and regulators alike want assurance that AI systems will not mishandle personal or critical data.

These efforts also align with a growing understanding that data governance is key to scaling AI responsibly. According to a global study by NTT Data, 35% of Chief AI Officers now say that enabling “private and sovereign AI” – AI that is built with protected data and runs in controlled environments – is their biggest barrier to adoption ([6]). Yet the same research indicates that organizations which make data and AI sovereignty a mission-critical priority are achieving up to 5 times higher returns on innovation and efficiency, and are 90% more likely to reach transformative AI outcomes ([7]) ([8]). In other words, the companies that proactively address data governance, privacy, and sovereignty are not just avoiding fines – they are positioning themselves for significantly better performance from their AI investments. In the coming months, we can expect more enterprises to follow the lead of these data-forward organizations, tightening data controls, investing in compliance-friendly architecture, and treating “data as law” as well as “data as an asset” in their AI strategies.