Making Sense of Semantics in the Age of AI

In 2023, Donald Farmer delivered a keynote titled “Making Sense of Semantics” during the BARC Retreat. It was a period in which AI was transitioning from experimentation and hype toward early forms of large-scale adoption, accompanied by growing concerns around scalability, governance, and real business value. At the same time, Farmer’s presentation exposed a fundamental issue: despite significant technological advances, many organizations continued to struggle with conflicting numbers, slow insights, and a fundamental lack of trust in their own data.

What made this session resonate for me was how closely it reflected the questions I hear increasingly often from clients. These are rarely questions about tools or platform choices, but about meaning. What is the real value of an ontology? Is it an actionable asset, or primarily a conceptual framework? How can such a model be applied concretely in reporting, integration, and analytics? And above all, what role does semantics play in a world where AI is becoming ever more central?

As organizations progress further along their AI and data-driven decision-making journeys, it is becoming clear that technology alone is not enough. Without shared meaning, data remains fragmented, insights remain contested, and AI remains unreliable. This is precisely where semantics —and ontologies as their explicit foundation— become essential to truly connect data, analytics, and AI with the business.

The Return of Semantic Layers as a Strategic Foundation

Semantic layers are not new. As Donald Farmer illustrates through early Business Objects patents (e.g. the concept of a “Universe”), the idea of abstracting raw data into business concepts emerged as early as the 1990s. At the time, the objective was clear: shield business users from technical complexity and allow them to work with familiar terms such as revenue, customer, asset, or location.

So why is it that semantic layers are back so prominently on today’s agenda?

The answer lies in the convergence of several forces that have fundamentally reshaped the data landscape. Organizations now operate dozens of data sources, analytics tools, and dashboards, each introducing its own interpretation of metrics and entities. Self-service analytics has empowered business users, but it has also amplified inconsistency when business concept definitions are not governed centrally. At the same time, data governance pressures have intensified: security, privacy, quality, and compliance can no longer be enforced tool by tool. Cloud data platforms have accelerated this shift by making virtualization more attractive than physical data duplication, while data mesh and domain-oriented thinking have emphasized shared metrics and data contracts over centralized databases.

In this context, semantic layers re-emerge as the only realistic abstraction layer capable of keeping complexity manageable. They provide a consistent business view across tools, domains, and platforms, enabling trusted analytics at scale. In short, organizations have not rediscovered semantics — they have reached a point where operating without them is simply no longer viable.

From Semantics to Ontology: Making Meaning Explicit

The renewed focus on semantic layers also reveals a deeper challenge. Semantics define meaning, but meaning alone is insufficient if it remains implicit or locked inside individual tools. As organizations scale and become more complex, semantics must be explicit, shared, and durable across systems, teams, and use cases — including AI.

This is where ontologies come into play. An ontology is the formal expression of semantics: it defines concepts, relationships, and rules in a clear and machine-readable way. While a semantic layer operationalizes meaning for reporting and analytics, an ontology anchors that meaning as a foundational organizational asset. It ensures that terms such as “asset,” “location,” or “performance metric” carry the same meaning regardless of whether they are used in reporting, integration, governance, or AI models.

In the age of AI, this distinction becomes critical. AI systems do not inherently understand business context; they simply amplify the semantics embedded in the data they consume. Without an explicit ontological foundation, AI runs the risk of reinforcing inconsistencies and producing confident yet unreliable outcomes. Ontologies therefore act as the missing bridge between data, analytics, and AI, grounding advanced capabilities in shared understanding rather than isolated interpretations.

Seen through this lens, semantic layers and ontologies are not competing concepts but complementary ones. Semantic layers make meaning usable; ontologies make meaning sustainable. Together, they form the backbone of a modern data strategy, enabling organizations to move from fragmented insights to shared understanding, to scalable, trusted business value.

Conclusion

As organizations continue to scale their data and AI ambitions, the real differentiator is no longer technology, but shared understanding. Semantic layers and ontologies provide the foundation needed to turn data into trusted insight, and AI into a reliable business capability. By making meaning explicit and durable, organizations can move beyond fragmented analytics toward sustainable, scalable value creation — where data, analytics, and AI truly work in service of the business.