In enterprise publishing, we often find that the content model and the publishing process are so tightly coupled that any change to one requires a total overhaul of the other. This “monolithic” approach is the primary reason many companies struggle to adopt new technologies like AI.
But here is what I find genuinely encouraging: the solution is not some distant architectural dream. It is a well-understood pattern, it is available today, and teams that adopt it are already seeing returns that go far beyond technical elegance. They are seeing speed, adaptability, and a kind of organizational confidence that comes from knowing your infrastructure can absorb whatever comes next.
Modular Orchestration for Future-Proofing
The core philosophy behind SiteFusion ProConsult is the strict separation of concerns. By using **MarkLogic** for content storage and **Camunda** for process orchestration, we allow your content to exist independently of the rules that govern its movement. This is more than just a technical preference; it is a business necessity in the age of AI.
When your workflow logic is separate, you can introduce AI agents to specific stages of your lifecycle — such as automated translation, metadata enrichment, or sentiment analysis — without modifying your DITA maps or XML schema. This modularity means that as AI models improve, you can swap out an older “task” in your BPMN model for a newer, more efficient one. No migration project. No six-month initiative. Just a configuration change that your team can execute with confidence.
And this is where the optimism comes in. For years, the structured content community has been building on a foundation — DITA, semantic XML, separation of content from presentation — that turns out to be almost perfectly suited for the AI era. The discipline that technical writing teams have practiced for two decades is now a strategic asset. Your carefully structured topics, your metadata, your taxonomy work: all of it is exactly what large language models need to deliver accurate, grounded, and trustworthy results.
Why This Moment Feels Different
We have been through waves of technology hype before, and the structured content world has earned its skepticism. But this moment has a different texture. Three things are converging at once.
First, AI capabilities have reached a threshold where they can perform meaningful work inside a publishing pipeline — not just generate rough drafts, but handle classification, validation, and cross-referencing at a level of accuracy that is genuinely useful in production.
Second, orchestration engines like Camunda have matured to the point where adding an AI step to a workflow is no more complex than adding any other automated task. The BPMN model gives you a visual, auditable map of exactly where AI is operating and what guardrails surround it. Nothing is hidden in code. Everything is visible, testable, and reversible.
Third — and this is perhaps the most important piece — the structured content community has spent twenty years building content that is machine-ready. DITA-based content is not unstructured text that needs to be scraped, cleaned, and hoped over. It is typed, classified, richly linked, and semantically marked up. When we talk about making content “RAG-ready,” DITA teams are not starting from zero. They are starting from a position of remarkable strength.
From Defense to Opportunity
The old framing around separation of concerns was often defensive: protect yourself from vendor lock-in, insulate yourself from change, hedge against the unknown. All of that remains true. But the new framing is about opportunity.
When your workflow layer is independent and visible, you can experiment. You can run a pilot where AI handles metadata tagging on a single content stream while everything else stays exactly as it is. You can measure the results. If it works, you expand. If it does not, you roll back. The cost of experimentation drops dramatically, and that changes the entire conversation with stakeholders.
We see this with our own customers. Teams that once spent months building the case for a single process change are now running multiple concurrent pilots, each scoped to a specific BPMN task, each delivering measurable results within weeks. The architecture makes the experimentation possible, and the experimentation builds organizational trust in the technology.
An Invitation: ConVEx 2026 in Pittsburgh
As the technical writing community gathers this week for the 28th annual ConVEx (formerly DITA North America) in Pittsburgh, I invite you to consider your own architecture. Is your workflow a hidden part of your code, or is it a visible map that your team can inspect, question, and optimize?
We will be at ConVEx all week, and we are hosting a Kitchen Talk today at 3:00 PM where we will walk through what this looks like in practice — live demonstrations of how a separated architecture handles real-world scenarios, from AI-assisted review cycles to automated compliance checking.
The future of technical documentation is not just about what we write. It is about how we orchestrate the flow of that information, who can see that orchestration, and how quickly we can adapt it. The good news is that the community gathered in Pittsburgh this week is better positioned for that future than almost any other discipline in enterprise technology.
Come find us. We would love to talk through what this could look like for your team.
