The situation
M&R ships a printed user manual with every machine and archives them for the ERP. Some of those manuals are twenty years old — written in a different era, structured around assumptions that no longer hold, and formatted in ways that make them painful to read and nearly impossible to reuse.
Migrating them into MadCap Flare meant starting from scratch every time. Review the PDF. Identify topics. Create each one individually. Rework the content into a structured template. Edit. Add image placeholders. Repeat — all of it squeezed between meetings with engineering, product management, and operations.
A full manual took an average of six weeks. Just getting topics created took about a week. For a 20-year-old manual with poor structure, it was longer. It wasn't sustainable.
The bottleneck
The problem wasn't effort — it was mechanical repetition. The information existed. It just needed to be pulled apart, reorganized, and rebuilt to a modern standard. That's exactly the kind of work a well-prompted API can do.
What was built
A Python script that takes a PDF user manual as input and outputs a set of .htm topic files structured for MadCap Flare — ready to import, complete with meta labels for SEO.
The Claude API handles the restructuring. It reviews the PDF, identifies topic types — operational tasks, maintenance tasks, troubleshooting — and reorganizes the content against a standard defining what belongs in each topic type. Content is rewritten to a 6th–8th grade reading level, formatted for LLM retrieval, and ANSI Z535.6 compliant safety messages are added where needed.
A manual that previously took several days to restructure now produces 30+ formatted starter topics in a few hours.
The output
Topics come out immediately importable into MadCap Flare with minimal formatting cleanup required. Content cleanup is still necessary — especially for older manuals where the source material is sparse or poorly structured — but that's expected. The tool doesn't replace editorial judgment. It eliminates the mechanical restructuring work so editorial judgment is all that's left.
What used to be days of topic creation is now a review session with an SME.
What this represents
The content standard, the API integration, the output validation against real manuals — this project sits at the intersection of technical writing, systems thinking, and software development. It's the kind of solution that only makes sense to someone who has done the manual work long enough to know exactly where the bottleneck is.