One of the best parts of my job is spending time with leaders who are putting AI at the center of their business strategy. Across industries and around the world, organizations are making significant investments — not because AI is interesting, but because they’re expecting real business results.
Those conversations all come back to the same question: What separates the companies getting measurable value from those still chasing promises?
Recently, I read Deloitte’s AI ROI Leaders study. The research reinforced much of what I’ve been seeing firsthand.
The highest-performing organizations don’t treat AI as another technology deployment. They approach it as a business transformation. They focus on growth, not just efficiency. Their CEOs are engaged. They invest in data, governance, and change management. And they prioritize a handful of high-value use cases instead of trying to boil the ocean.
I agree with all of those conclusions, but I’d add one more.
The organizations making the most progress balance long-term ambition with short-term execution.
Too many AI programs begin with a sweeping enterprise vision and assume value will naturally follow. The reality is that enterprise transformation takes time. If the business doesn’t see measurable outcomes along the way, momentum fades, priorities shift, and confidence erodes.
The most successful organizations think differently. They establish an enterprise architecture that can scale, while delivering tangible business outcomes one use case at a time. Every successful deployment solves a real problem, builds trust across the organization, and creates the foundation for what’s next.
That’s especially true when building an enterprise semantic layer and ontology. The end goal is a shared understanding of enterprise data across the business, but organizations shouldn’t have to wait years to realize value. They should see measurable impact at every stage of the journey.
The Foundation Determines the Outcome
Another takeaway from Deloitte’s research stood out to me.
The companies seeing the strongest AI returns aren’t succeeding because they found a better model. They’re succeeding because they invested in the fundamentals first.
They prepared their data. They established governance. They created business context. They redesigned processes where necessary.
In other words, they built a foundation AI could actually work with.
AI by itself doesn’t create business value.
AI creates value when it’s grounded in trusted, connected, business-ready data.
That’s why semantic layers and knowledge graphs have become so important. Without that context, AI is left making educated guesses based on probabilities. With it, AI can understand how information fits together, reason across complex relationships, and produce answers the business can trust.
Of course, not every semantic strategy delivers the same result.
Over the years, I’ve come to believe there are four principles that matter most.
Four Principles for Enterprise Semantics
1. Connect to Enterprise Data Wherever It Lives
I’ve yet to meet a large enterprise whose critical data lives in one place.
Organizations run on ERPs, CRMs, operational systems, cloud platforms, data lakes, SaaS applications, and countless other repositories. That’s simply the reality of modern IT.
A semantic layer has to embrace that reality — not fight it.
If your architecture depends on moving every piece of enterprise data into a single repository before AI can use it, you’ve created another bottleneck. The semantic layer should connect to data where it already exists and present it as one coherent view.
2. Build on Open Standards
Technology will continue to evolve. Vendors will change. AI platforms certainly will.
Your enterprise knowledge shouldn’t have to change with them.
Open standards give organizations flexibility, interoperability, and control over one of their most valuable assets: their knowledge. They ensure today’s AI investments remain valuable regardless of tomorrow’s technology choices.
3. Virtualize Before You Materialize
For years, enterprise architecture centered on copying and consolidating data.
That approach doesn’t scale well in an AI world.
Modern semantic architectures should let AI understand data where it already resides, materializing data only when there’s a clear operational or performance reason to do so.
The objective isn’t creating more copies of enterprise data.
It’s creating a better understanding of it.
4. Build on an RDF Knowledge Graph
The semantic layer provides the business context.
An RDF knowledge graph provides the reasoning capability.
It represents business entities, relationships, concepts, and rules in a form machines can understand, and just as importantly, infer new relationships that weren’t explicitly modeled.
As enterprise AI moves beyond answering questions toward reasoning, planning, and autonomous decision support, that capability becomes increasingly important.
Separating Real Enterprise AI from Marketing
Right now, nearly every technology vendor claims to provide context for AI.
The term has become almost as ubiquitous as AI itself.
But providing context labels and delivering enterprise understanding at scale aren’t the same thing.
I think of it like physical fitness.
Most people understand what it takes to get into great shape. Far fewer consistently do the work required to achieve it.
Enterprise semantics is no different.
Closing the AI-Ready Data Gap
Most organizations recognize that AI needs trusted business context. The gap between recognizing that need and building for it is significant.
Accenture’s AI-ready data research found that 64% of organizations have moved beyond pilots into production across multiple functions or launched coordinated enterprise-wide initiatives. Yet only 7% have reached the level of data readiness required to scale. At the same time, 72% lack trusted, high-quality, governed data suitable for advanced AI, and more than 80% have delayed, limited, or changed AI initiatives because of data-related risks.
Closing that gap takes more than better data quality and governance. AI-ready data must bring structured and unstructured information together, capture business context and domain knowledge, define the semantic meaning of entities, concepts and relationships, and remain fresh and accessible to AI agents and applications.
That is precisely what an enterprise semantic layer and knowledge graph are designed to provide: a shared, machine-readable understanding of the business that connects information with the context AI needs to reason and act with confidence.
The payoff is measurable. Accenture’s “Data Reinventors,” the 7% that have achieved AI-ready data, report stronger productivity, decision quality, customer experience and risk reduction, along with up to 1.6x greater profit margin improvement than industry peers.
That’s the challenge we’ve been focused on solving at Stardog from the beginning.
Our mission isn’t simply to make AI smarter. It’s to help organizations give AI the context, understanding, and trusted knowledge it needs to deliver measurable business outcomes.
