What challenges did Johnson & Johnson MedTech face before AI implementation?

Manual inspection and documentation of surgical trays was complex, time-consuming, and error-prone. Johnson & Johnson MedTech needed better visibility and traceability across a large surgical set supply chain. The business needed a scalable way to identify, cleanly catalog, and validate more than 400 instruments per set across tens of thousands of trays.

What was the impact of the AI solution?

Inspection and documentation time was reduced by over 40%. Technician onboarding time dropped from eight to 12 months to just three months. The solution supported lower headcount during the pandemic while maintaining output. The system delivers a one in 10,000 real-world error rate. The rollout expanded from a Queensland pilot to live operations across Australia, New Zealand, and Japan.

Source

GCPProductionEvidence: Medium50/100

SAVI transforms global surgical instrument tracking with Google Cloud

Use case typeComputer vision checkoutUpdated Jun 13, 2026

Powered by Vertex AI, SAVI (Semi Automated Vision Inspection) is transforming surgical instrument identification and cataloging to reduce missed instruments and manual inspection workload. Johnson & Johnson MedTech worked with Max Kelsen and Google Cloud to build a system that manages tens of thousands of individual devices and their tray characteristics for surgical sets used by surgeons. The solution was piloted in a Queensland distribution center serving over 100 hospitals and then rolled out across Australia, New Zealand, and Japan.

Organization: Johnson & Johnson MedTech
Industry: Healthcare
Location: Australia
Published: January 2023

Reported outcomes

3 months

technician onboarding timeTime & speed

−40%inspection and documentation time reduction1 in 10000error rate

Strategic outcomes

Market & geographic expansionExpanded operations across Asia-Pacific marketsRisk & complianceImproved surgical tray traceability and validationNew product / capabilityAutomated surgical instrument inspection and catalogingSpeed & agilityStreamlined inspection and documentation workflow

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Inspection and documentation time reduction: 40% decrease

Google Cloud BlogJan 6, 2023Blog postExplicit claimMedium evidence strength

reduced the time Johnson & Johnson MedTech needs to check and document inspections ... by over 40%

Normalized claim

Technician onboarding time: 3 months decrease

Google Cloud BlogJan 6, 2023Blog postExplicit claimMedium evidence strength

from eight to 12 months down to just three months

Normalized claim

Error rate: 1 in 10000 decrease

Google Cloud BlogJan 6, 2023Blog postExplicit claimMedium evidence strength

delivering a one in 10,000 real-world error rate

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Johnson & Johnson MedTech
Provider: GCP
Maturity: Production
Linked source: Google Cloud Blog

The solution was piloted in a Queensland distribution center serving over 100 hospitals and then rolled out across Australia, New Zealand, and Japan

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 3 of 4

1Computer Vision
2Operational Efficiency
3Supply Chain Optimization

Manual inspection and documentation of surgical trays was complex, time-consuming, and error-prone.
Johnson & Johnson MedTech needed better visibility and traceability across a large surgical set supply chain.
The business needed a scalable way to identify, cleanly catalog, and validate more than 400 instruments per set across tens of thousands of trays.

A tablet and web application captures tray images through an API endpoint hosted on Google Cloud.
Vertex AI Workbench is used for data exploration and modeling.
Kubeflow training pipelines run on Google Kubernetes Engine to produce machine learning models for specific surgical instrument sets.
Hundreds of machine learning models are served with Kubeflow model serving on GKE for low-latency inference.
The system checks whether instruments are correctly placed, dirty, broken, bent, or otherwise not fit for purpose, then returns results to the tablet for user action.

Technologies

Vertex AI Vertex AI Workbench Kubeflow Google Kubernetes Engine GKE

Inspection and documentation time was reduced by over 40%.
Technician onboarding time dropped from eight to 12 months to just three months.
The solution supported lower headcount during the pandemic while maintaining output.
The system delivers a one in 10,000 real-world error rate.
The rollout expanded from a Queensland pilot to live operations across Australia, New Zealand, and Japan.

Architecture

A tablet and web app captures tray images and sends them via API on Google Cloud. Vertex AI Workbench is used for data exploration and modeling, Kubeflow training pipelines run on GKE, and hundreds of machine learning models are hosted with Kubeflow model serving on GKE for low-latency instrument identification and tray validation.

Implementation partners1

Max Kelsen

Sources & evidence1

Evidence: Medium50/100Evidence strength