Scaling AI-powered inspection

SINCE ITS beta launch in June 2025, more than 100 customers worldwide have used OneVision to accelerate AI-powered vision development and deployment, with many progressing from single-line applications to multi-site rollouts in days instead of months. This momentum reflects a broader shift as manufacturers move beyond isolated AI pilots toward connected, enterprise-wide inspection strategies.

"AI vision has long delivered value, but scaling it across operations has remained a barrier,” said Matt Moschner, president and CEO of Cognex. "Manufacturers encounter recurring challenges, from fragmented workflows to models that don’t adapt across environments. OneVision addresses this by unifying the simplicity of the edge with the scalability of the cloud, helping organisations move from isolated pilots to consistent, enterprise-wide deployment."

Scalable AI vision

OneVision addresses a persistent challenge in industrial AI: deploying advanced vision applications at enterprise scale without adding complexity or slowing production. This introduces a cloud-to-edge architecture, where AI models are trained, managed, and governed in the cloud, while inspection runs at the edge on Cognex vision systems for real-time, reliable execution. Customers can now centrally manage the entire AI lifecycle, from collecting and labeling production images to refining models, and deploy updates consistently across global fleets of devices. OneVision is optimised to work with Cognex’s latest systems, including the In-Sight 3900 and In-Sight 6900.

“While OneVision leverages the cloud for development and management, runtime inspection remains fully edge-based,” said Reto Wyss, vice president of vision engineering at Cognex. “Once a model is deployed, no connectivity to the cloud is required. Production images stay local and latency is a non-issue.”

By centralising model development and management, OneVision helps manufacturers standardise inspection processes across sites, reduce duplication of work across teams, reduce scaling costs by up to 50%, and maintain version control and consistency across deployments.

Standardising AI inspection for global scale

One customer to reap the benefits of OneVision is Schneider Electric. Using the Advanced AI tools in OneVision, engineers label images with a high degree of granularity and details as well as refine ROI and retrain models rapidly One of the most powerful capabilities is the ability to train models efficiently – even starting with minimal image sets – and refine them iteratively within a collaborative environment. 

The platform has also eliminated an ongoing hardware challenge. Previously, advanced AI inspection often required PCs equipped with high-performance graphics cards. With OneVision, that requirement disappeared. This not only reduced direct hardware costs but simplified system architecture and maintenance. 

As with any cloud platform in a global enterprise, security validation was critical in evaluating OneVision. Schneider Electric conducted a comprehensive security assessment of the cloud-based solution. Security and engineering teams from Cognex supported the process by completing third-party security questionnaires and providing necessary certifications. The platform met Schneider Electric’s most demanding global standards.

The broader testing and validation process spanned approximately six months. Much of the effort was concentrated at the beginning, where teams focused on training, validation, and ensuring operational reliability before full-scale deployment.

Despite the rigorous process, the ease of use and ergonomic interface of OneVision accelerated adoption. Engineers found that new inspection solutions could be developed and integrated 30% faster compared to previous approaches. 

“OneVision allowed us to develop and validate AI inspection standards centrally and then deploy those same models across our worldwide operations,” said Christophe Ernis, smart operation manager, product power division. “That approach helped us double yield, dramatically reduce false rejects, and reduce our dependence on specialised vision expertise. Most importantly, it gives us a repeatable way to scale best practices reliably across our factories.”

www.cognex.com