Structured Content: The Missing Piece in AI Architecture

You can’t scroll through LinkedIn these days without encountering elaborate diagrams of AI architecture. They are everywhere. With spooky colors and black backgrounds, they illustrate how large language models (LLMs), retrieval systems, AI agents, orchestration tools, and governance mechanisms work together to deliver business results.

I’ve been looking at a number of these models recently. Some are remarkably detailed, with layers devoted to everything from data management to security and human oversight. But something stands out: Many acknowledge the importance of enterprise data and knowledge, yet few explicitly address structured content or the professionals who manage it.

That’s a consequential omission. For decades, technical communicators, content strategists, and information architects have developed methods to organize, classify, maintain, and reuse information. These methods have renewed relevance as organizations build trusted AI systems that depend on reliable organizational knowledge.

Overlook that relevance at your own risk.

Structured Content: Established Principles, New Relevance for AI

Structured content has been extensively defined, described, and standardized by international bodies and leading institutions. It is information that is organized into modular, discrete, self-contained components that are:

• Consistently architected for human and machine readability
• Identifiable
• Semantically tagged (without regard to formatting)
• Version-controlled
• Reusable
• Made relative and relevant with component-level metadata

Standardized information architecture helps both people and machines understand what these components contain, how they fit together, and where they apply.

Consider a company’s product documentation. Instead of creating separate manuals for every product variation and user role, a structured-content approach lets writers develop reusable components, such as procedures, warnings, and product descriptions. These components can be classified, maintained, conditionalized, and assembled for different products, audiences, or delivery channels—usually with digital tools and increasingly with agentic AI tools.

The principles behind structured content aren’t new. Technologies and standards such as XML and DITA have supported structured authoring and content management for decades. But their relevance extends beyond technical documentation.

Structured content can provide a critical foundation for trusted AI systems that depend on enterprise knowledge. Knowledge engineering expert Michael Iantosca argues in his paper “The Agentic Documentation Age Has Exposed the Limits of Markdown and Docs-As-Code” that reliable, reusable enterprise knowledge must have a managed semantic layer. “For any AI system expected to deliver accurate, traceable, context-aware results at enterprise scale, it is rapidly becoming a requirement.”

But why does structured content matter when today’s AI systems can process vast amounts of unstructured information?

AI Can Read Content, But Can It Reliably Use It?

LLMs can process enormous amounts of unstructured information, including documents, websites, emails, and other text-based content. But processing information doesn’t guarantee that they can use it effectively to meet users’ needs. The challenge lies in determining whether the information is accurate, current, relevant, and appropriate for a particular situation.

Consider an AI assistant helping a customer troubleshoot a product. The company’s knowledge base contains instructions for several product versions, including procedures that have been revised or replaced. The AI assistant might retrieve information that appears relevant to the customer’s question. But can it reliably distinguish the current procedure from an obsolete one? Can it determine which instructions apply to the customer’s particular product?

Structured content and robust management software, such as a component content management system (CCMS), can help resolve these uncertainties. Explicit metadata, version information, and relationships among content components provide context that AI systems can use to identify applicable information. Content governance helps ensure that the underlying information remains accurate and current.

“The future is not ‘XML instead of AI.’ It is structured, governed semantics underneath AI – so the model has less to infer, less to hallucinate, and far more to know.”

Michael Iantosca

Evidence of the potential benefits comes from a tcworld article, “iiRDS for Trusted AI.” The authors describe an implementation involving technical documentation for mining equipment in which semantically enriched content improved AI retrieval and the identification of safety-critical information compared with less structured approaches.

Of course, structured content alone cannot guarantee trusted AI output. System design, retrieval methods, validation, and human oversight still matter. But when organizations make the meaning and context of their information explicit, they give AI systems a stronger foundation for producing reliable results.

The Overlooked Role of Content Professionals in AI

Structured content doesn’t create, organize, or maintain itself. Behind the information architecture are people with specialized expertise, many of whom have spent their careers helping organizations make information accurate, accessible, and reusable:

• Technical communicators
• Content strategists
• Information architects
• Knowledge engineers
• Content managers

In his article, “The Knowledge Layer CIOs Keep Overlooking,” content strategy expert Scott Abel argues that organizations that employ these professionals already have expertise directly relevant to the challenges AI teams are trying to solve: organizing, governing, and delivering trustworthy information.

Yet organizations don’t always include these professionals when planning AI initiatives. AI architects and developers may focus on models, data pipelines, orchestration, and automation. But they often fail to consult the people who understand how the organization’s content is structured, maintained, and used. Abel faults leaders who see structured content tools as operational, not strategic.

That failure of vision can lead to:

• Proliferation of inconsistencies in content deliverables
• Increased localization costs
• Unnecessary grunt work
• Duplicated effort
• Missed opportunities
• Unnecessary risks, especially in highly regulated industries

There’s another consequence that deserves attention: the potential loss of institutional knowledge.

Much of an organization’s product, process, and compliance knowledge resides in the experience of the people who create and manage its content. They understand not only why information is organized the way it is and how content components relate to one another, but also which sources are authoritative, and what must change when products, policies, or processes evolve.

Not all of that knowledge is captured in documentation or software.

When experienced content professionals retire, leave, or are excluded from AI transformation efforts, organizations risk losing knowledge that may be difficult and costly to reconstruct. (I raised a related concern in last month’s blog post about documenting AI proofs of concept, where I emphasized preserving lessons learned before they disappear.)

Organizations developing AI capabilities should involve content professionals early, recognize their expertise as part of the enterprise knowledge infrastructure, and ensure that their knowledge is preserved and transferred. Their contributions can strengthen both AI reliability and the long-term value of the organization’s information.

Making Structured Content Part of AI Architecture

As organizations continue to design increasingly sophisticated AI architectures, they need to look beyond models, agents, and orchestration tools. They should also examine the quality and structure of the information on which their AI systems depend.

That means recognizing structured content as a strategic asset, assessing existing content architecture and governance practices, and bringing content professionals into AI planning from the beginning.

The expertise and methods needed to strengthen an evolving AI architecture may already exist within the organization. The challenge is to recognize their value, preserve them, and put them to work.

The next time you encounter an elaborate AI architecture diagram, ask yourself: Where’s the structured content? And where are the people who understand it?

Photo by Clark Van Der Beken on Unsplash.

Disclosure: The author used AI tools during outlining and initial drafting work.