The AI revolution

The AI content landscape has transformed dramatically since 2022. What began as simple text generation has evolved into sophisticated AI systems capable of reasoning through structured content and completing complex tasks independently.

This rapid progression tells the story. In 2022, we experimented with AI for basic text generation - helpful but limited. By 2024, systems like Tridion Docs Genius integrated AI assistants into content portals through retrieval-augmented generation, creating trustable chat interfaces that could accurately answer questions based on your documentation.

Now in 2025, we're seeing truly agentic systems emerge. These AI agents can reason through structured content, understand relationships between components, and complete complex documentation tasks with minimal human guidance.

For organizations using DITA and component content management systems like Tridion Docs, this progression is transformative. The structured, semantic nature of DITA content provides an ideal foundation for these next-generation AI systems.

The careful structure technical writers have built into documentation systems has become a competitive advantage. Systems originally developed for human knowledge transfer are proving remarkably well-suited for human-AI collaboration.

This post explores how Tridion Docs with agentic AI uses DITA's task-based approach can deliver truly intelligent assistance - and what this means for technical documentation professionals.

The rise of Agentic AI

Technical documentation has always faced a fundamental challenge - bridging the gap between information and action. Users don't want documentation - they want to accomplish tasks. This disconnect is what agentic AI now addresses when paired with structured content systems like Tridion Docs.

For content teams using Tridion Docs, this shift fundamentally changes content delivery. Traditional search and early AI focused on finding and presenting information. Agentic systems actively help users accomplish tasks using that information.

This distinction matters because it solves several persistent documentation challenges:

• First, it addresses the focus problem. Users often struggle to identify which parts of documentation apply to their specific situation. An agentic system can maintain context across interactions, understanding the user's particular environment and goals rather than treating each query as isolated.
• Second, it moves beyond the limitations of retrieval-augmented generation (RAG). While RAG improved AI responses by grounding them in accurate documentation, it still primarily answered questions rather than guiding complete processes. Agentic AI uses documentation as a foundation for action, not just information.
• Third, it creates a bridge to non-public data sources. Many organizations maintain critical information across disconnected systems - product databases, customer records, configuration settings - that traditional documentation can't easily incorporate. Agentic systems can connect these disparate sources when helping users complete tasks.

What makes an AI system truly “agentic” is its ability to demonstrate autonomy, maintain goal-orientation across interactions, and execute planning rather than simply responding to prompts. These capabilities transform how users interact with their content.

Consider a manufacturing technician needing to reconfigure equipment after a line change. With traditional documentation, they'd need to locate relevant procedures, interpret which sections apply to their specific equipment model, and manually track their progress. With an agentic system connected to Tridion Docs' content, they engage with an assistant that:

• Understands their specific configuration requirements
• Accesses relevant task topics from Tridion Docs DXD
• Connects to equipment databases to verify compatibility
• Guides them through the process with contextual awareness
• Adapts instructions based on their responses
• Maintains focus on the overall goal despite interruptions

For organizations using DITA, this approach uses existing investments in structured content. The task-oriented architecture of DITA - with clearly defined goals, prerequisites, steps, and results - provides the perfect foundation for agentic systems to understand not just what information exists, but how that information helps users accomplish real-world objectives.

The challenge now is connecting these AI capabilities with enterprise content systems in a standardized way that maintains security and accuracy - a bridge that specifications like Anthropic's Model Context Protocol are beginning to build.

Anthropic’s MCP - building bridges to legacy systems

The promise of agentic AI is compelling, but technical documentation teams face a practical challenge: how to connect advanced AI models to the structured content already managed in products like Tridion Docs. This is where Anthropic's Model Context Protocol (MCP) becomes a critical enabling technology.

MCP represents a significant advancement in how AI systems interact with external data sources and tools. At its essence, MCP provides a standardized framework that allows AI models to request specific information from external systems, process and reason over that information within context, take actions through external tools when needed, and maintain awareness of available capabilities.

For technical documentation environments, this capability transforms how we work. Rather than requiring complete migration of content into AI-native formats or building custom integrations for each use case, MCP offers a flexible middleware approach that respects and uses existing investments.

Understanding MCP's architecture

The Model Context Protocol operates through a client-server architecture with well-defined interactions:

• Context requests
  The AI agent identifies when it needs additional information and makes a structured request for specific data using the protocol's request-response pattern.
• Tool use
  The agent can invoke predefined functions to interact with external systems through the protocol's messaging capabilities.

The protocol uses JSON-RPC 2.0 for communication, with clear message types for requests, responses, notifications, and errors. This structured approach suits technical documentation systems particularly well, as they already organize information in purpose-specific components.

MCP handles the communication flow between LLM applications (hosts) and servers that provide context and tools but leaves implementation details like persistence to the specific applications using the protocol.

For further details see:

https://modelcontextprotocol.io

The legacy system challenge

Most enterprise organizations have spent years building comprehensive documentation repositories. These systems represent massive investments not just in content creation, but in taxonomies and metadata structures, review workflows, compliance processes, integration with product lifecycle management, and translation pipelines.

Replacing these systems isn't feasible or desirable. Yet such organizations still want to use the latest AI capabilities to improve content search and user experience.

MCP addresses this challenge by providing a standardized interface layer that can connect to existing content repositories without disrupting established workflows. For Tridion Docs, this means continuing to author in DITA, manage content through familiar processes, and use the existing delivery pipelines - while simultaneously enabling AI-powered experiences.

Practical implementation approaches

Implementing MCP with legacy documentation systems typically follows one of several patterns:

• API-based integration
  Creating connectors that allow AI agents to query content repositories through existing APIs. For Tridion Docs, this might use the Content Delivery API to retrieve published DITA topics.
• Search-based retrieval
  Using enterprise search capabilities as an intermediary, allowing agents to discover relevant content through semantic search before processing it for user interactions.
• Cached knowledge bases
  Periodically extracting structured content into AI-optimized formats that maintain the semantic richness of DITA while enabling faster retrieval.
• Hybrid approaches
  Combining real-time API access for current information with cached knowledge for common scenarios, optimizing for both accuracy and performance.

The beauty of MCP is that it abstracts these implementation details from the AI model itself. The agent simply requests information about a particular topic or task, and the MCP framework handles the complexities of retrieving that information from your Tridion Docs environment.

Beyond simple retrieval

What makes MCP particularly valuable for technical documentation is that it goes beyond simple RAG (retrieval-augmented generation) approaches. Rather than just pulling content and generating responses, MCP enables true tool use - allowing agents to query specific metadata about documentation components; filter content based on product versions, user roles, or environments; navigate complex documentation structures based on user needs; and potentially even trigger workflows within the CCMS itself.

This capability transforms technical documentation from a static resource into an interactive system that can actively assist users through complex tasks - precisely what DITA's task-oriented approach was designed to enable.

With this understanding of MCP's capabilities, let's explore how we could implement this technology within Tridion Docs DXD's Content Delivery system to create truly intelligent documentation experiences.

Connecting Tridion DXD to Agentic systems

Architectural Overview

The Tridion Docs Dynamic Experience Delivery (DXD) platform is a high-performance content delivery framework optimized for structured documentation.

It provides an API-first architecture built on GraphQL endpoints and provides advanced search and retrieval of Tridion Docs Topics. It was built to support customers of all sizes, including those with enterprise documentation needs.

Topics are stored within an OpenSearch cluster, together with a range of metadata that enables end-users to search for topics based upon term-based matches, natural language questions, and existing topic context. It comes with a sophisticated recommendation engine for locating the most relevant content based upon an existing topic. Metadata, such as Product, Version, Audience and Goal form an integral part of the hybrid retrieval process.

MCP integration points

The most effective integration points uses DXD's existing capabilities:

• GraphQL Content API
  DXD's GraphQL API offers a structured way to retrieve DITA topics and their relationships. This API becomes the primary conduit through which MCP requests can access documentation content.
• Search integration
  DXD's OpenSearch capabilities provide an ideal entry point for natural language queries from agentic systems. When an agent needs to locate relevant content based on user intent, it can use the existing search APIs rather than building a separate retrieval mechanism.
• Metadata services
  The rich metadata layer in DXD - including product hierarchies, versions, task and audience definitions - provides crucial context for agents. MCP implementations can query this metadata to understand content applicability and filter results appropriately.
• Topic relationships
  DITA's inherent linking structure, exposed through links within the DXD APIs, allows agents to understand relationships between concepts, tasks, and reference materials. This enables more sophisticated reasoning about documentation structure.

Task-focused topics

Topics in DITA contain several elements that make them particularly valuable for AI agents:

• Clear goal statements
  Topics typically begin with a purpose statement that explicitly defines what the user will accomplish. This maps directly to user intent in conversational interactions with AI.
• Prerequisites
  Topics can include prerequisites, helping agents understand what conditions must be met before a task can be performed. This enables more intelligent assistance when users lack necessary preparations.
• Ordered steps
  The sequential, numbered steps in DITA topics provide agents with a clear procedure to guide users through. Each step's discrete nature makes it easy for agents to track progress and provide contextual help at specific points.
• Results
  Task often include expected outcomes, allowing agents to confirm successful completion and troubleshoot when results differ.
• Examples
  The examples included in many topics give agents concrete illustrations to share with users when abstract instructions prove challenging.

Implementation overview

Spring AI provides a robust foundation for building agentic AI applications within the Java ecosystem. At its core, Spring AI offers abstractions that simplify integration with various AI models and vector databases, allowing developers to focus on business logic rather than provider-specific implementations.

For our Tridion Docs DXD content delivery API, Spring AI is particularly valuable because it provides a consistent interface across multiple AI providers while maintaining access to model-specific features when needed. This flexibility allows us to switch between different AI services without significant code changes, protecting our investment as the AI landscape evolves.

The Model Context Protocol (MCP) implementation in Spring AI fits naturally with our existing GraphQL API, creating a bridge between structured content delivery and agentic AI capabilities. By using Spring AI's MCP server capabilities, we can expose our content delivery APIs as tools and resources that AI agents can discover and utilize at scale, turning our documentation system into an interactive knowledge base.

Tridion Docs DXD is already designed for high-performance content retrieval, making it ideal for real-time AI interactions where response speed is critical. This existing strength pairs exceptionally well with Spring AI's MCP server implementation, allowing us to deliver AI-enhanced content experiences without sacrificing the performance our users expect.

We’d use the WebFlux SSE transport option for our MCP server, which provides further advantages for our use case. The reactive programming model of WebFlux aligns perfectly with DXD's high-performance retrieval capabilities, allowing us to handle many concurrent AI agent connections without blocking threads. This non-blocking approach means our server can maintain responsiveness even under heavy load, which is essential for production AI applications. Additionally, Server-Sent Events (SSE) provides an efficient one-way communication channel that's ideal for streaming AI responses as they're generated, creating a more responsive user experience.

For further details see:

https://docs.spring.io/spring-ai/reference/api/mcp/mcp-overview.html

Example scenarios

Below are some example scenarios, and the MCP tools that an agent could use.

Complete a task

• Agent searches for a topic via the search-topic DXD MCP tool
• Extracts prerequisites, steps, and expected results via the extract-content DXD MCP tool
• Guides user through process with contextual awareness
• Adapts to user feedback during execution

Troubleshooting an issue

• Agent identifies potential issue from user description
• Retrieves relevant troubleshooting topics via the search-topic DXD MCP tool
• Cross-references with known issues database via a get-issues MCP tool
• Presents progressive resolution options

Context-aware documentation

• Agent identifies current user context (product, version, environment)
• Retrieves topic relevant to immediate need via the get-recommendations DXD MCP tool
• Proactively offers related recommendations based on DXD's recommendation engine
• Maintains awareness of user's progression through documentation

Implementation challenges

There are some implementation challenges that would need to be addressed when building such a service.

Content currency

• Ensuring agents access the most recent documentation when content updates occur
• Tridion Docs DXD provides access to the latest published versions of content

Response latency

• Optimizing retrieval operations to meet user expectations for real-time interactions
• Pre-fetching likely-needed content based on conversation context

Metadata utilization

• Translating user context into appropriate metadata filters
• Balancing precision (exact metadata matches) with recall (semantically relevant content)

Error handling

• Gracefully managing scenarios where content is unavailable or access restricted
• Developing fallback strategies for incomplete information
• Transparently communicating limitations to users when documentation gaps exist

Example implementation

See this post for an example implementation using Spring AI. 

Business benefits

Implementing agentic AI with Tridion Docs promises tangible business value across multiple dimensions. Organizations that connect their structured DITA content to intelligent agents can expect significant returns through improved user experiences, operational efficiencies, and strategic advantages.

Enhanced user experience

The most immediate benefit would be the transformation of the documentation experience. Traditional interfaces require users to navigate, search, and synthesize information themselves. With agentic systems using DITA content, users could:

• Express needs in natural language instead of searching for precise terms
• Receive information tailored to their specific context
• Experience complex procedures as interactive conversations rather than static text

Reduced support costs

When users can successfully complete tasks through AI-guided documentation, they should require less human support:

• Fewer support tickets for documented procedures
• Faster solution identification when tickets do occur
• More first-line resolutions without escalation to product specialists

Improved content discovery

The structured nature of DITA content should provide significant advantages for agentic retrieval:

• Task topics map directly to user intent
• Conditional processing facilitates personalized delivery
• Relationship tables and content references help agents understand connections
• Metadata and taxonomies improve retrieval accuracy

These structural advantages would translate to users finding relevant information more quickly, increased documentation engagement, and previously “buried” content becoming discoverable through conversational interfaces.

Content quality insights

Agentic documentation systems could provide unprecedented insight into content effectiveness. By analyzing where users struggle despite AI assistance, content teams might identify:

• Missing information in topics
• Confusing terminology or instructions
• Common user questions not addressed
• Logical gaps in task procedures

Competitive advantage

Beyond cost savings and efficiency gains, agentic documentation systems should transform technical content from a business expense to a competitive differentiator. When structured DITA content connects with modern agentic AI, organizations could experience:

• Superior self-service experiences with customers solving problems independently
• Faster onboarding and reduced time-to-value for new customers
• Broader product adoption as novice users successfully navigate complex features
• Increased feature exploration as users gain confidence in finding assistance
• Higher customer satisfaction leading to improved retention

Preparing for implementation

To position for these advancements, organizations should:

• Audit content structure to ensure DITA implementations follow task-oriented best practices
• Train technical communicators on AI capabilities and limitations
• Identify high-value use cases where agentic assistance would most benefit customers
• Evaluate delivery infrastructure for compatibility with emerging AI integration standards

The organizations likely to gain the most significant advantages are those who have maintained discipline in their structured content practices. The careful work of technical writers in creating well-structured, task-focused documentation should pay unexpected dividends in the age of agentic AI.

Conclusion

The integration of agentic AI with Tridion Docs represents a decisive moment in the evolution of technical documentation. What we've explored isn't just a technological curiosity - it's a practical way to transform how organizations create, manage, and deliver technical information.

The journey from traditional documentation to intelligent, interactive assistance isn't as daunting as it might initially appear. By using Anthropic's Model Context Protocol as the bridge between the Tridion Docs infrastructure and modern AI capabilities, customers could begin realizing benefits quickly while building toward a more comprehensive vision.

In summary

• Structure creates advantage
  The structured, task-oriented nature of DITA content in Tridion Docs provides an ideal foundation for agentic AI systems. The investment in well-organized technical documentation is now more valuable than ever.
• Start with delivery
  Connecting agents to the existing content delivery layer offers immediate benefits with minimal disruption to existing workflows.
• Think beyond retrieval
  The most powerful implementations go beyond simply finding information - they help users accomplish real tasks through interactive guidance.

The technical documentation community has always been at the forefront of structured information management. Now, as AI capabilities mature, that structured approach is showing its value. The semantic richness and task-oriented design of DITA content provides exactly what intelligent systems need to deliver truly helpful assistance.