Today
Every new manual starts with copying from old PDFs, and the mistakes of the past travel along. Thirty years of manuals sit as scans, PDFs and text files in folders and archives, some from an authoring system that no longer exists. Which text applies to which assembly, and which version is right, only long-serving colleagues know.
An example
Conveyra Fördertechnik has 640 operating and maintenance manuals for overhead cranes and hoists, written over 30 years, many only as scans. demi splits them into 21,000 sections, recognises about 2,300 distinct modules among them and finds 38 places where the inspection intervals for the same wire rope contradict each other, quarterly versus half-yearly. The writing team settles the contradictions with engineering and builds the manual for the new BK 10 t overhead crane largely from approved modules.
Your data, wherever it lives
Your old manuals sit today as scans in the archive, as PDFs on drives and in the DMS, as text files in project folders and partly in an authoring system that no longer exists. demi reads, understands and links every chapter and module to product and assembly in the PLM, with your rights. The AI splits in days what would take a writing team many months, and the team decides which version applies.
Where the data lives today
- Scans & paper
- PDFs
- Folders & drives
- PLM
- DMS & archive
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
Data sources in this case: scanned and digital manuals from folders and the DMS, plus the PLM with products, assemblies and bills of materials. demi replaces nothing and turns every manual into chapters and modules, each linked to product, assembly, information type and origin. After approval the modules go into your authoring system.
- 2 · PermitPermission Layer
The technical writing team releases modules; safety and warning texts are released only by those authorised for them. Manuals for special machines stay with the project team, and every release is recorded with version and person.
- 3 · ActIntelligence Layer
The agent recognises chapter structure, task steps, warnings and figures, in scans too, and classifies every module according to iiRDS. It groups identical and near-identical texts, shows contradictions in values and intervals with both sources and proposes one version per group. It checks technical values exactly against the engineering data in the PLM.
How it runs
- 1
demi takes in old manuals from folders, archive and DMS, plain scans included.
- 2
The agent splits them into chapters and modules: task, description, warning, figure.
- 3
Every module gets product, assembly and information type according to iiRDS.
- 4
Duplicate and contradictory texts are grouped, values checked against the PLM.
- 5
The writing team decides per group which version applies and releases the modules.
What changes
New manuals from checked modules instead of copies
Old contradictions surface before they travel on
Decades of knowledge stay usable
What does it bring you?
Check it with your own numbers
Every starting value is an assumption you can change. The calculation sits right below, step by step.
Your numbers
Result
+230 hfreed per year
Assumptions and calculation
demi splits the manual into classified modules and flags duplicates; you decide what applies.
8 h → 4.8 h per case
How we calculate
6 manuals × 8 h × 40% × 1.0 (manual) × 12 months = 230 h
Value per year, before demi costs
Freed per year+230 h
Assumption: labour cost of €49.50 per hour in German manufacturing in 2025 (Destatis), plus workplace, IT and overheads.
= Value per year, before demi costs
+€14,950
gained per year
230 h ÷ 1,760 h per role = 0.1 full-time roles
An estimate, not a promise. The starting values are assumptions for an industrial company with 150 to 500 people.
Combine with more cases and get the packageQuestions about this use case
Does this work with scanned manuals too?
Yes. The agent reads scans with text, tables and figures and recognises the chapter structure even without a digital original. Where a page is hard to read, it flags it instead of guessing.
What does classification according to iiRDS mean?
iiRDS is an open standard for metadata in technical documentation. Every module gets product, assembly, information type and life cycle phase, such as installation or maintenance. Writers, service and portals then find the right module, whichever manual it came from.
How does demi handle contradictory texts in old manuals?
It shows them instead of silently picking one. For every group of near-identical modules you see the versions side by side, differences marked, with manual and year. For values such as inspection intervals or tightening torques it also shows what the PLM says; the writing team decides which version applies.
Do we need an authoring system for this?
No, but it helps. If you have one, the modules go there after approval. Without one you use them directly in demi, for example for the operating manual per variant.
