# Publications from the practice.

Four documents — three practitioner guides and one collection of engagement records. Each names tools and standards, takes positions, and is written to be read by a working architect or retrieval engineer — not skimmed for marketing copy.

## What's in this section.

Most firms publish "resources" that read as competitively-bid content marketing — generic capability descriptions, problem-cost-opportunity frames, related-content sidebars. The four documents below are not that. They name tools and standards, take positions, and are written to be useful to a working DITA architect, conversion engineer, retrieval specialist, or publishing engineer — practitioners who can already disagree with us if they want to.

Three guides codify the methodology: AI readiness as a content engineering problem, migration as recovery rather than transcription, manual publishing as a risk surface that CI/CD reduces. The fourth is a record set — three anonymized engagement case studies that demonstrate the work the guides describe. Read the guide that maps to what you're trying to do; read the case studies to see what the same approach has produced.

## The four publications

1. **01**  
   ### [Engagement records.](https://www.extense.co/company/our-work-and-case-studies)  
   Three engagements from the practice — Enterprise DITA Training Curriculum, AI-Ready Content Pipeline, Custom DITA-OT Publishing Automation. Anonymized; verifiable on request under NDA.

Inside  
   * * *  
     - Three case records, each with hero summary card and full artifact ledger  
     - Engagement metadata: duration, team composition, scope, status  
     - Standards & tooling stack named per case  
     - Decisions and trade-offs: the architectural choices and what was rejected

2. **02**  
   ### [Practitioner guide](https://www.extense.co/resources/ai-readiness)  
   Retrieval precision is upstream of model selection. The seven dimensions of content engineering that determine whether an AI pipeline returns the right answer or the closest keyword match — with a 0-14 scorecard, an interpretation rubric, and worked examples of metadata, chunking, and vocabulary in practice.
   
   Inside  
   * * *  
     - Seven-dimension scorecard with 0 / 1 / 2 rubric  
     - Interpretation: what an aggregate 0-14 score means  
     - Five-phase preparation roadmap with deliverables  
     - Worked examples: metadata filtering, section-aware chunking, controlled vocabulary  
     - JSONL output schema — the artifacts that ship to a vector store

3. **03**  
   ### [Practitioner guide](https://www.extense.co/resources/migration-playbook)  
   A playbook for moving content from unstructured legacy formats — Word, FrameMaker, MadCap, RoboHelp, AuthorIT — into typed, validated DITA. Patterns drawn from two million pages of migration work across federal, defense, life sciences, and enterprise commercial engagements.
   
   Inside  
   * * *  
     - Migration lifecycle: Audit → Map → Convert → Refine → Publish  
     - Five disciplines: content audit, IA, pilot migration, batch conversion, dedup & enrichment  
     - Thirteen source formats covered, named explicitly  
     - Four field rules with the failure modes that justified them

4. **04**  
   ### [Practitioner guide](https://www.extense.co/resources/automation-guide)  
   A guide to CI/CD for documentation — containerized builds, an eight-stage validated pipeline, parallel multi-format output, and chatbot-ready JSONL as a first-class artifact. Real code blocks, real platform recommendations, real validation gates.
   
   Inside  
   * * *  
     - Eight-stage pipeline: commit to production, with validation integrated into every stage  
     - Containerization principle + production Dockerfile  
     - Four output formats — HTML5, branded PDF, JSONL, additional — with build-time data  
     - Four CI/CD platforms — GitHub Actions, Azure DevOps, Jenkins, GitLab CI  
     - Three custom DITA-OT plugins, versioned and maintained as Git artifacts

## Sample Content Assessment

Submit a 20-page sample. We'll return conversion feasibility, content recovery rate, and engineering effort within two business days. The analysis is the basis for any further engagement, with no obligation to proceed.