NAIT — Curriculum Analysis Agent (DocGen fit assessment)¶
Status: candidate — assessed, not built. Verdict: Phase-1 gap analysis is not a DocGen-template fit; the Phase-2 Content Feeder is.
What NAIT does¶
NAIT writes apprenticeship textbooks — ILMs (Individual Learning Modules) — for trades like Electrical. The province (AIT — Apprenticeship & Industry Training) updates the official curriculum guidelines. When they change, NAIT's trade specialists (SMEs) must read thousands of pages of old textbook vs new guidelines and decide, line by line, what's new / changed / removed / unchanged. Manual today: ~4,400 hours, ~4 months per cycle. Goal: an AI tool that automates the comparison (~4,400 → ~150 hours).
- Phase 1 (5 wks, Electrical): prototype that compares old textbook + old/new guidelines → a gap-analysis table (each outcome tagged Net New / Change / No Change / Removal, + difficulty level, + AIT-clause & ILM-section citations) → Excel; scored on recall/precision vs NAIT's human answer key.
- Phase 2 (15 wks): more trades; real AltaForge UI with per-row Accept/Reject/Annotate; a Content Feeder that surfaces draft text per gap and produces a culled textbook.
Is DocGen a fit?¶
No, for the Phase-1 gap analysis. DocGen is a template → prose/sectioned document generator. The gap analysis is a multi-step classification + alignment pipeline whose deliverable is a per-row grid (Excel), reviewed per-row, scored on classification recall/precision.
| DocGen | NAIT gap analysis | |
|---|---|---|
| Output | prose document | classification grid (.xlsx) |
| Core task | write grounded prose | align + classify across 2 docs |
| Hard part | faithful prose | classification recall / precision |
| Review UX | per-section edit/regen | per-row accept/reject/annotate |
Shared DNA (why the question is fair): document ingestion, mandatory citations/grounding, no fabrication, eval-vs-gold. Those are AltaForge platform values — the foundations — not the document-template model.
Where DocGen genuinely fits: the Phase-2 Content Feeder — take a legacy ILM + the gap classifications and generate a culled/reorganized textbook (drop Removal sections, relocate Moved). That's real document generation; the gap analysis is the classifier that feeds it.
Recommendation¶
- Build Phase 1 as its own AltaForge solution type (classification/alignment pipeline + review grid), reusing the gather/grounding/eval foundations — not the document-template machinery.
- Reserve a DocGen template for the Content Feeder in Phase 2.