Medtronic · Intelligent extraction of engineering drawings and spare parts

A Evidence level

Material ID search revolutionized with Microsoft GenAI + IBM Consulting

Publisher: IBM · Case of the official customer of the manufacturer · Direct sources at the case level

Claim origin: IBM Disclosures with Clients · Independent verification: No · Accessed: 2026-09-19

Evidence level measures whether a source can be located and reviewed; it does not mean vendor-reported claims were independently audited.

Deep case | Key delivery chain is substantially documented 12 / 12

Business context2 / 2

Transformation workflow2 / 2

Technical workflow2 / 2

Human roles & governance2 / 2

Measured outcomes2 / 2

Source traceability2 / 2

All six dimensions meet the current completeness threshold.

Medtronic deconstructs engineering drawings into OCR, RAG validation, structural warehousing and expert review, enabling AI to move beyond “reading” to entering material identification and supply chain decision-making processes.

Business problem

The large number and complex structure of engineering drawings for medical instrument products, and the reliance on manual reading and checking for key raw materials and spare parts identification, both slow product development and supply chain analysis, can easily lead to omissions due to differences in format and profile.

Solution

Medtronic Global Operations and Supply Chain Analysis Teams, Vendor Management Teams and IBM build custom programs: Azure Document Intelligence first makes OCR and layout extracts, Azure OpenAI-supported RAG validates material information in the engineering context, key attributes are written into the database and search applications are available to business personnel.

Technical architecture & production workflow

Step 01

Complex project drawings and design documents

→

Step 02

Azure Document Intelligence OCR and Page Parsing

→

Step 03

Context of RAG Retrieval Project

→

Step 04

Azure OpenAI generates and validates candidate properties

→

Step 05

Key materials and parts number entered into database

→

Step 06

Search applications and expert anomaly review

Key technology & infrastructure components

Azure Document IntelligenceAzure OpenAIRAGAzure Blob StorageProperties DatabaseExpert review

Human roles & accountability

AI completes map identification, information extraction and candidate generation; supply chain and engineering specialists validate key attributes, handle low confidence exceptions, and are responsible for material and design decision-making.

FDE delivery actions

Define key extraction fields with supply chain and engineering teams
Combining OCR determinative Resolution with RAG semantic understanding
Measuring the accuracy of materials and spare parts numbers, respectively
Design of database and search portal entry into business processes
Leave the low-confidence results to the experts.

Reusable delivery patterns

Complex document AI should split OCR, semantic understanding and manual review
Different fields should be assessed separately rather than only reporting the total accuracy rate
The result must be extracted into a searchable data structure
High-risk engineering information requires confidence and exceptional mechanisms

Business outcomes & delivery results

The IBM case revealed that the accuracy of critical material information extraction was up to 90 per cent, the accuracy of spare parts number extraction was over 85 per cent, significantly reducing manual verification and saving several hours of processing; and the overall cost savings were not disclosed.

Evidence boundaries & verification notes

90 per cent and 85 per cent, respectively, of the corresponding material information and spare parts numbers were extracted.
The public page does not give samples, error distribution and external evaluation.

Primary source: Material ID search revolutionized with Microsoft GenAI + IBM Consulting ↗