Why Your Content Strategy Needs to Get Smart: Key Insights from Noz Urbina

The buzz around AI and content is everywhere, but are we actually doing it right? In a fascinating conversation on the Content Matters podcast, Noz Urbina,  a globally recognized leader in the field of content strategy and customer experience, breaks down what it really takes to build AI-powered content systems that actually work—and why most organizations are getting it wrong.

The Foundation: Understanding Semantic Content Models

Before we can make AI truly useful for content, we need to understand what a semantic content model actually is. Think of it as creating a blueprint that makes sense to both humans and machines.

Instead of just dumping information into database tables, a semantic model gives everything human-readable labels and shows how different pieces of content relate to each other. For example, when creating content about an actor, you'd define attributes like "name," "filmography," "biography," and "origin," specifying which elements are required and how they connect.

This isn't just about organization—it's about creating content that can actually answer complex questions. When your content is properly structured, AI can handle queries like "show me European films featuring this actor" instead of just spitting out generic responses.

Knowledge Graphs: The Smart Database Behind the Magic

Here's where things get interesting. A knowledge graph isn't your typical database with rows and columns. It's more like a web of interconnected concepts where the relationships themselves have meaning.

The real power comes from these connections. If an actor changes their name, every reference across your entire content ecosystem updates automatically because you've changed the central relationship, not individual links. It's like having a smart assistant that keeps track of how everything connects.

Why LLMs Aren't Enough (And What They're Missing)

Large Language Models like ChatGPT are impressive, but they have a fundamental limitation: they work on statistical probability, not true understanding. They're great at giving you popular, general answers but struggle with specific, niche questions that require deep domain knowledge.

Imagine asking, "How do my specific 15 product modules work together given the versions I'm currently running?" An LLM will likely give you a generic response because it's not a common query. A knowledge graph, however, can trace those exact relationships and give you the precise answer you need.

This is where the real magic happens—when you combine both systems.

The Perfect Partnership: LLMs + Knowledge Graphs

The most effective AI content systems use LLMs and knowledge graphs as a team:

The LLM handles the conversation: It understands your question, processing natural language, and presenting answers in a friendly, readable way.

The knowledge graph provides the expertise: It stores accurate, domain-specific information and complex relationships that the LLM can tap into.

When you ask a question, the LLM translates it for the knowledge graph, which finds the accurate answer, then the LLM translates that back into natural language. You get the best of both worlds: conversational AI with factual accuracy.

Getting Started: Domain Modeling First

If you're thinking about building an AI-powered content system, Urbina's advice is clear: start with domain modeling, not AI implementation.

This means thoroughly mapping out the fundamental concepts in your field and how they relate to each other. In the film industry, that might be understanding how actors, directors, films, and years connect. In your business, it could be products, features, customers, and use cases.

This foundational work isn't glamorous, but it's essential. You need this structured understanding before AI can do anything meaningful with your content.

The Costly Mistakes Everyone's Making

Organizations are rushing into AI without doing the groundwork, leading to predictable failures:

• Treating LLMs like databases: Feeding messy PDFs into AI and expecting accurate answers

• Layering AI on top of chaos: Using more AI to fix problems caused by poor content organization

• Building unnecessary custom models: Investing heavily in proprietary AI when better general models already exist

• Believing in magic solutions: Expecting AI to automatically boost productivity without proper processes or human oversight

The harsh reality? AI can't fix bad content—it just makes the problems faster and more expensive.

Why Content Strategists Are Having Their Moment

For years, content strategists have been advocating for treating content as a structured business asset. AI has finally made this approach not just nice-to-have, but essential.

Content strategists who understand machine-readable content (sometimes called "content engineers") are now in prime position. They can:

• Build the structured foundation that makes AI actually useful

• Design workflows where humans focus on creative, high-value work while AI handles the repetitive tasks

• Ensure accuracy and prevent the "hallucinations" that plague poorly implemented AI systems

Urbina said it's a bit of a "told-you-so" moment for content professionals who've been preaching structure and strategy all along.

The Bigger Picture: Ethics and Human Agency

Beyond the technical challenges, Urbina raises important questions about AI's broader impact. When AI systems are optimized for engagement and built with embedded biases, they can subtly influence how we think and what we believe.

The concern isn't that AI is inherently evil, but that it's a powerful tool that can amplify existing problems if we're not careful. The risk of "truth collapse"—where algorithmically generated content becomes indistinguishable from factual information—is real and growing.

The Path Forward

The key takeaway from this conversation is that successful AI implementation isn't about the AI itself—it's about the content foundation you build first. Organizations that invest in proper content strategy, semantic modeling, and knowledge graph development will see real returns. Those that try to shortcut the process will likely waste time and money on solutions that don't actually solve their problems.

Ready to Dive Deeper?

This summary only scratches the surface of the insights shared in this episode of the Content Matters podcast. For the full conversation, including more detailed examples and practical implementation advice, be sure to check out the complete episode. You'll get a much richer understanding of how to build AI-powered content systems that actually deliver value for your organization.

‍