Artificial intelligence is moving quickly from experimentation into day-to-day operations. Customer support, product documentation, internal knowledge systems, and decision support are all being reshaped by large language models (LLMs).
What many organizations are discovering just as quickly is that model capability is not the limiting factor. Content is.
Teams are deploying AI on top of content environments that were never designed for consistency, reuse, or scale. The result is uneven: answers that conflict, gaps in context, and outputs that require verification before they can be trusted.
Organizations that get past this stage are not choosing different models. They are addressing the content layer those models depend on.
The hidden cost of content chaos.
The pressure on enterprise content is not new. AI is simply exposing it.
Research from Adobe shows that 95% of organizations struggle to measure the ROI of their technical content. At the same time, the Adobe State of Work report highlights that 87% of organizations struggle to manage content across its lifecycle, while 62% report difficulty finding and sharing information.
In practice, that fragmentation looks familiar:
- Duplicate content across teams and systems
- Conflicting terminology and inconsistent structure
- Limited visibility into ownership and lifecycle
- Gaps between pre-sale and post-sale content-led experiences
These issues already slow down publishing, increase costs, and complicate governance. When AI is introduced, they do something more significant: they reduce confidence in every output.
In many organizations, this shows up when support teams, product teams, and documentation teams provide different answers to the same question.
LLMs surface what already exists. When content is inconsistent, that inconsistency becomes visible at scale.
AI trust runs on structured data.
Modern AI systems act better when provided structured data. NVIDIA CEO Jensen Huang stated during the NVIDIA GTC 2026 Keynote, “Structured data is the foundation of trustworthy AI.”
That framing makes the dependency clear.
Every answer an AI system produces is assembled from underlying content. If that content is incomplete, duplicated, ambiguous, or flat-out wrong, the output reflects it. If it is structured, consistent, vetted, human-approved, and properly governed, the output improves accordingly.
Huang expands on this further saying, “This is the structured data… the ground truth of business. This is the ground truth of enterprise computing.”
Content is now moving into the same role.
Structured content already delivers measurable results.
The value of structured content is not theoretical. It has been measured in operational and financial terms.
An IDC study on Adobe Experience Manager Guides found that organizations adopting structured content approaches achieved:
- $3.8 million in average annual benefits per organization
- 287% return on investment over three years
- A 13.9-month payback period
- 17% improvement in technical writing productivity
- 16% reduction in duplicate content effort
- $5.8 million in additional annual revenue impact
These gains were realized even before AI became a primary driver of content strategy.
They come from a set of capabilities that are straightforward but difficult to achieve without structure:
- Reuse instead of duplication
- Consistency across channels
- Modular content that can be recombined
- Governance that maintains accuracy over time
Forrester research supports the same pattern. 69% of organizations report that structured, XML-based content enables reuse and syndication, and more than 80% say a component content management system (CCMS) reduces regulatory, financial, and reputational risk.
A practical example illustrates how this translates into AI-readiness. In its Adobe DITAWORLD 2024 session, Ernst & Young explained that “all page content is stored in DITA XML format… [and] eventually, it will feed into LLMs to enable generative AI chat experiences”.
The connection is direct: Structure created for reuse and governance becomes the same structure that supports AI.
Where Darwin Information Typing Architecture (DITA) fits — and why LLMs benefit.
This is where DITA moves from background detail to strategic relevance.
DITA is not simply a documentation standard. It is a system designed to enforce the exact characteristics AI systems require:
- Content broken into modular, portable, self-contained topics
- Consistent structure across all information types
- Built-in metadata and taxonomy
- Reuse that eliminates duplication and conflict
- Governance that ensures accuracy over time
In other words, DITA operationalizes structured content at scale.
This matters because LLMs do not 'understand' documents. They retrieve, rank, and assemble content fragments. The quality of those fragments determines the quality of the output.
DITA produces fragments that are:
- Predictable in structure
- Consistent in terminology
- Context-rich through metadata
- Maintained as a single source of truth
This is why, in practical terms, LLMs perform better when fed content derived from DITA-based systems. Not because the model prefers XML, but because the underlying content has already been normalized, structured, and governed.
The earlier point from NVIDIA becomes more concrete in this context. If “structured data is the foundation of trustworthy AI,” then DITA is one of the most effective ways enterprises create that foundation for content.
For organizations already using DITA, this creates an advantage:
- Content is already modular and reusable.
- Metadata already exists to support retrieval.
- Governance is already in place to maintain trust.
AI does not require a reinvention of content. It requires a content model that was designed correctly in the first place.
DITA provides that model.
Why structure matters more with AI.
AI does not introduce new content challenges. It amplifies existing ones.
The same foundations that improve publishing performance also determine whether AI outputs are usable:
- Consistency: Single sourcing reduces conflicting answers.
- Modularity: Topic-based content aligns with how retrieval systems work.
- Metadata: Tagging provides context for ranking and personalization.
- Governance: Version control ensures accuracy over time.
Without these, organizations spend time cleaning and reconciling content before AI can be trusted. That effort grows quickly as use cases expand.
With them, AI-ready content is not a separate initiative. It is an extension of how content is already created and managed.
Content as infrastructure.
The shift underway is less about documentation and more about operating models.
Content is moving toward an infrastructure role, with a lifecycle that resembles a supply chain:
- Creation and standardization
- Governance and validation
- Storage and reuse
- Transformation into multiple outputs
- Delivery to both humans and machines
When this model is in place, a single source can support multiple outcomes — PDFs, web experiences, knowledge bases, and AI-driven interactions — without introducing inconsistency.
This mirrors how data platforms evolved, where raw inputs are transformed into curated datasets that can be reused across analytics and operations.
A feedback loop that compounds value.
As AI becomes embedded in workflows, content and AI begin to reinforce each other.
- Content informs AI outputs
- AI interactions highlight gaps and improvements in content
- Updates are made once and reused everywhere
- Outputs improve over time
Organizations with structured content can move through this cycle quickly and with control. Those without it often see the opposite: duplication increases, inconsistencies spread, and governance becomes harder.
Rethinking where advantage comes from.
Much of the current discussion around AI focuses on models and infrastructure. Those elements matter, but they are widely accessible.
What differentiates organizations over time is how well they manage the inputs those systems rely on.
Structured content does not replace AI. It makes it dependable. DITA is one of the clearest implementations of that principle in practice.
Organizations that invest in structured, governed, and reusable content are doing more than improving documentation. They are creating a foundation that supports accurate, scalable, and consistent AI-driven experiences.
Moving from chaos to advantage.
AI outcomes are closely tied to content quality.
Organizations that continue to treat content as a by-product will find it difficult to scale AI beyond isolated use cases.
Organizations that treat content as infrastructure can move faster, reduce risk, and deliver more consistent experiences.
In an AI-driven enterprise, content is no longer just something that supports operations.
It becomes part of how operations run.
Where Adobe Experience Manager Guides fits.
This is where a platform approach matters. Adobe Experience Manager Guides operationalizes DITA at enterprise scale — bringing structured authoring, component reuse, metadata management, and governance into a single system of record.
With Experience Manager Guides, organizations can:
- Author in DITA with enforced structure and consistency.
- Manage components with versioning and workflow-driven governance.
- Apply taxonomy and metadata that improve findability and retrieval.
- Publish to multiple channels from a single source.
- Generate AI-ready outputs (clean HTML/JSON, chunked content, metadata-enriched corpora).
The result is not just faster publishing. It is a reliable content foundation that can feed search, self-service, and AI experiences with the same governed source.
As AI use cases expand, platforms like Adobe Experience Manager Guides turn structured content from a documentation practice into an enterprise capability — one that supports accurate, consistent, and scalable outcomes across every channel, including AI.
Saibal Bhattacharjee is the Director of Product Marketing for the Digital Advertising, Learning, and Publishing Business Unit at Adobe.
Saibal has been with Adobe for 16 years, and is currently in charge of global GTM and business strategy for a diverse product portfolio in Adobe — ranging from market-leading cloud-native component content management system (Adobe Experience Manager Guides), advertising and subscription monetization products for connected multiscreen TV platforms (Adobe Pass) to content authoring and publishing desktop apps (Adobe FrameMaker, Adobe RoboHelp).
With more than 21 years of experience in the technology sector, Saibal is a high-impact marketing, strategy, and product executive with a passion for tackling the most complex challenges in enterprise software and turning solutions into scalable works of enterprise-grade art. He has successfully built, mentored, and managed global GTM teams spanning India, US, UK, Germany, and Japan for more than a decade. Saibal holds a BE degree from Jadavpur University, Kolkata, and an MBA degree from the Faculty of Management Studies, University of Delhi.
