ARC Champions Radar: AI and ML Solutions from Automation Suppliers

Executive Overview

Industrial-grade AI isn’t just any AI, it’s the kind that’s been through the industrial wringer and comes out ready for the factory floor, clipboard in hand and safety boots on. As coined and discussed during this year’s ARC European Industry Forum in Spain, industrial-grade AI must be robust, explainable (no black box) reliable, and able to survive not only real-time applications, time-series data and edge deployments, but also the scrutiny of every engineer in the plant.

It’s not enough to be smart; the experts at the AI sessions underlined, industrial-grade AI must be... well, it wasn’t easy to define. Session members came up with an abbreviation all could agree on and laugh about. Industrial AI has to be USSR: Understandable, Safe, Secure, and Reliable.

Yes, in a twist of history, the USSR is back, but this time, it’s not about geopolitics, it’s about making sure your industrial AI doesn’t start a five-year plan for downtime. If your AI solutions can explain themselves, keep your data safe, survive a cyberattack, and still show up for work on Monday, congratulations, you may have industrial-grade AI. Let's put the winking to one side for now. The above definition is not accurate enough for the OT world, even though it’s catchy. Let’s go with this one:

Industrial-grade AI refers to artificial intelligence systems engineered for deployment in industrial environments, including real-time specific applications. Industrial-grade AI requirements include:

Robustness and reliability

Explainability and trustworthiness

Domain-specific intelligence

Edge Deployability

OT integration

Compliance with industrial standards, guidelines and certifications

Industrial-grade AI and ML are increasingly integral to digital transformation initiatives, supporting manufacturers and end users in aligning advanced technologies with operational priorities and digitalization goals. Industrial AI is distinguished by its robustness, explainability, domain-specific intelligence, edge deployability, and compliance with industrial standards.

Key requirements include reliability, transparency, integration with operational technology (OT), and the ability to withstand rigorous industrial scrutiny. The report highlights that successful industrial AI must be understandable, safe, secure, and reliable, ensuring it delivers measurable value without introducing new risks.