Mark van Berkel is Co-Founder and CTO of Schema App, specializing in ontology and semantic data for search, AI and agentic experiences.
Almost every executive I speak with has an AI initiative underway. The initiatives vary. Some are rolling out copilots to improve productivity while others are testing AI-powered customer experiences. But they all have one thing in common: They’re focused on how AI can drive business outcomes.
Far fewer are asking what AI needs to do those things well.
After enough of these conversations, I’ve noticed a pattern across enterprises. Teams are putting significant effort into evaluating models and testing use cases, but the quality of the results is often limited by something much more fundamental: AI can only work with the information it understands, and for most organizations, that understanding is incomplete.
This isn’t because enterprises lack data; it’s quite the opposite. They often have thousands, sometimes millions, of web pages. The challenge is that this content was created for people, not machines.
Humans can naturally connect ideas and build context, but AI can’t (yet). It has to infer those relationships from whatever signals are available, and inference is never as reliable as understanding.
Although models and agent interfaces will continue to change, the need for a shared, machine-readable understanding of your business won’t. That’s why building a semantic data layer belongs at the start of your AI strategy.
Building Understanding For AI
For years, the goal of digital strategy was to get people to your website. Organizations invested heavily in content because search engines rewarded businesses that published useful, authoritative information. When people could find you, they could learn about and evaluate your products, and ultimately become customers.
Much of that is still true. But AI now answers user questions about your brand instead of simply pointing them to your website. Those answers are assembled from multiple sources, often before a customer has ever visited your website.
The important thing to note is that people are more often interacting with AI’s understanding of your business before interacting with your business itself. When every page, product, expert and service exists in isolation, AI must connect the dots on its own. Sometimes it succeeds, but sometimes it overlooks important context or draws the wrong conclusions.
That’s why you need to make your organization’s knowledge easy for machines to understand, connect and reuse. But how do you do that?
The Value Of Structured Data Has Grown Beyond SEO
Creating that shared understanding starts with structured data. Many organizations still view structured data as an SEO tactic. This is understandable, as its value was historically most visible in search.
In today’s search landscape, that same structured data supports a much broader range of technologies. Search engines still rely on it, but so do LLMs, internal search systems and an emerging ecosystem of AI agents. They all depend on clear, consistent information to understand an organization’s entities and the relationships that connect them.
These broader use cases change the return on investment of structured data. Defining your business in a machine-readable way creates a single source of truth about your brand that can be reused wherever AI needs to understand your organization.
The Semantic Data Layer Of Your AI Strategy
A semantic data layer creates a shared, machine-readable representation of your business. Tags on pages are necessary, but they aren’t the model itself. Where structured data are the building blocks, a semantic data layer is the shared model those tags are supposed to express: a governed, reusable representation of your entire business. It defines your products, services, areas of expertise and the relationships between them, giving every AI application access to the same trusted, brand-controlled understanding.
Many organizations are investing in copilots, AI agents and customer service automation as separate initiatives. Without a shared data foundation, every project rebuilds its own map of the business: what you offer, how it relates to other offerings, who your experts are and so on. That’s the same application-centric pattern that produced decades of integration debt, now with a chat interface. The modeling work gets repeated again and again across different teams and platforms.
That’s why building a semantic data layer should be the starting point of your enterprise AI strategy. Create it once, govern it centrally and make it available wherever AI is deployed. By doing so, you strengthen every subsequent investment by giving each AI system a reliable foundation to build upon.
Conclusion
New models, agents and standards such as model context protocol (MCP) and NLWeb will continue to change how machines connect to enterprise content. What won’t change is their dependence on reliable, shared meaning.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
