Parts Catalog Compiler
Agentic automation that generates interactive catalogs from 3D models and BOMs with 70% or greater reduction in authoring time, 90% or greater auto-link coverage across parts and documents, and technician search that supports UTID serial and natural language with under 2 seconds query latency.
Challenge
Building interactive parts catalogs is slow and manual. Teams juggle Illustrator masking, hierarchy building, linkages to BOMs and documents, and repeated validations across disconnected tools in PLM/PDM, DAM, spreadsheets, 3D models, and manuals. Existing assets are underused. Technicians lack UTID serial lookup and natural language search across parts and procedures, causing longer field time and rework. Each model release demands many tool hops and dozens of hours to produce and validate the catalog.
The objective: Generate interactive catalogs from 3D models and BOMs with ≥70% reduction in authoring time, ≥90% auto-link coverage across parts and documents, and technician search that supports UTID serial and natural language with ≤2 seconds query latency.
Solution: How AIP changed the operating model
Learning and setup
Powered by the Aftermarket Intelligence Platform, the agentic solution applied its vision model for segmentation, detection, and OCR, NLP retrieval for semantic search, a policy model for authoring and validation rules, and predictive ranking for disambiguation. Training data came from 3D CAD and exploded view renders, historical Illustrator files and masking rules, BOM spreadsheets, service manuals and procedures, UTID serial mapping tables, technician clickstreams, and annotated screenshots. This enabled the AI agent to recognize UTIDs and serials, model and revision, part IDs and BOM lines, 3D meshes and layers, SVG region masks, document IDs, and localization metadata.
Workflow orchestration
The AI agent analyzes authoring requests and change events, extracts model identifiers, and routes work across ingest, mapping, validation, and publishing. It navigates PLM/PDM, DMS/DAM, ERP, catalog CMS, render services, and search indexers, mirroring a content specialist workflow. Orchestration branches on confidence and policy—for example, when part-to-BOM matching is below threshold or duplicate names are detected—while versioning assets, enforcing checks, logging audits, and routing low-confidence items to human queues.
Execution and resolution
The AI agent executes end to end: import 3D and BOM, align parts to BOM entities, auto-generate the assembly hierarchy, create region masks and SVG hotspots, detect and label parts in images, link parts to procedures and documents, build semantic and UTID serial indexes, run coverage and naming validations, and publish to the catalog CMS. Runs complete in minutes with updates posted across connected systems. Exceptions such as missing CAD, mismatched IDs, ambiguous links, or duplicate SKUs are escalated to subject matter experts with diffs and suggested fixes attached.
