← Back to siteSolutions / DITA documentation blueprint中文EnglishРусский
G BUSINESS×Content Engineering
Proposal · v1.0
Oct 2026 · Client decision deck
DITA conversion of technical documentation · Process and acceptance design

Word in,
structured knowledge out.

R&D writes the data, AI converts the format, engineers do the final refinement. One pipeline turns 5,000 legacy Word files into reusable, conditionally publishable content assets.Write once, and it also powers review, PDF and AI Q&A。

View the full process

A documentation workflow that R&D will actually use

06Stages · 6 acceptance gates
05Roles · one shared pipeline
01DITA master · single source of truth
5,000Legacy documents · single entry point
01 / Process

Six stages, five swimlanes,one closed delivery loop

Six steps, one loop. The platform is replaceable — swap in XDocs and roles and acceptance criteria stay the same. Tap“Demo mode” on the right to walk through each stage.

Role actions Deliverables G1Acceptance gates Review loop
Scroll sideways to view
WHO / Roles

Swimlanes

5 LANES
01Intake

Intake

G1 · Intake
02Convert

Convert

G2 · Convert
03Refine

Refine

G3 · Quality
04Review

Review

G4 · Review
05Resolve

Resolve

G5 · Closure
06Publish

Publish

G6 · Delivery
LANE 01 R&D R&D / SME Content source
R1 · AuthorINPUT

Write principles and timing in the Word template; attach port / register data files.

Output · Template + data
R2 · ReviewREVIEW

Return changes using track-changes and comments only; no restyling or restructuring.

Output · Review comments
LANE 02 Platform Platform Replaceable · XDocs
P1 · IntakeINTAKE

Submit to GitLab intake intake — an MR is created automatically.

Output · Intake rules
P2 · ValidateCI

Run schema validation and generate an HTML preview.

Output · CI config
P3 · GovernGOVERN

Manage conditions and the reuse library.

Output · Condition model
P4 · ExportEXPORT

Export the review Word file with baseline SHA and bookmarks.

Output · Review exporter
P5 · BaselineBASELINE

After merging, tag a new baseline and start the next round.

Output · Version record
P6 · PublishPUBLISH

Publish PDF / WebHelp by condition.

Output · Publish sample
LANE 03 AI AI Services Conversion & QA
A1 · ConvertCONVERT

Split topics by style, convert data to tables, produce a confidence report.

Output · Converter
A2 · QAQA

Check terminology, style and structure; suggest reuse opportunities.

Output · QA rule set
A3 · DiffDIFF

Compare against the baseline and produce change suggestions and a review report.

Output · Comment extractor
LANE 04 Documentation engineer Documentation The last mile
D1 · RefineREFINE

Fix low-score items in Oxygen and attach conditions and reuse.

Output · Edit volume
D2 · ResolveRESOLVE

Review and merge the MR, accepting or rejecting each comment.

Output · Resolution report
LANE 05 FAE / Customer Customer Content consumer
F1 · ConsumeCONSUME

AI Q&A searches published content and links answers back to the source topic.

Output · AI search
Review loop G5 New baseline ↺ G4 Re-export

Every round starts from DITA masterRe-export; Word is only the review surface,is never the master version。

A

Single source of truth

Word is only the input and review surface. DITA is the single master; PDF, web and AI retrieval all come from it.

B

Interchangeable platform

Choosing the XDocs route? The swimlanes and roles stay the same. Test conditional management, the review portal and version baselines.

C

Legacy backlog, batched

The 5,000 legacy Word files enter at stage 1. Triage and de-duplicate first, run 50 documents to measure effort, then schedule.

01 / 06 G1 · Intake

R&D submits using the template; the platform creates an MR automatically.