Structured content as the foundation for AI
Agnès and Anne-Marie are measured about AI. They do not see it as a threat. They see it as an accelerator for specific kinds of work: drafting first versions, translating input from non-English speakers, and helping subject-matter experts contribute without needing to know DITA or structured authoring conventions.
Their current constraint is translation. In medical devices, fully automated AI translation is not yet an accepted approach. Validation requirements and organizational confidence are not there yet. But they expect that to change, and when it does, they want their content to be ready.
AI readiness grounded in structure, not hype
"There is nothing better than structured documents for AI," Agnès says. The logic is straightforward. AI systems perform better when trained on organized, consistent, well-governed content. A team that has spent years building structured reusable topics has, in effect, already done much of the preparation work that AI adoption requires.
Review Space and Draft Space (Collective Spaces)
The team has not yet adopted Draft Space or Review Space. The reason is honest and familiar: they are always chasing deadlines, and carving out the setup time for a new workflow has not been possible. There is also a specific barrier with Draft Space: many potential contributors are not writing in English, and they do not know topic-based writing conventions.
This is where they see AI playing a near-term role. An AI layer that could take a subject-matter expert’s input, in any language, and transform it into correctly formatted DITA-ready content before it enters the system would open reviewer and contributor participation significantly. That is the workflow they want to build.
The deeper lesson: invest in the foundation before you need to scale
Both Agnès and Anne-Marie offer the same advice to peers starting out: understand how your writing choices affect translation before you write the first topic and invest the time in setting up the system correctly before you need it to perform under pressure.
That advice is easy to ignore when teams are under immediate delivery pressure. It is harder to ignore when you consider that a documentation model built on copy-and-paste and manual updates cannot survive a serious regulatory change without significant rework. Structured content is the foundation. Build it once. Maintain it. Let it compound.