As autonomous AI moves onto plant floors, mines, and construction sites, professional engineers are facing a governance question their predecessors never had to ask: who is in charge when the network goes down?
Steve Yates, a professional engineer with 35 years of experience in edge computing and industrial systems, argues that much of today’s industrial AI has quietly reversed the architecture that made industrial control reliable. For decades, PLCs governed processes locally — interlocks, logic, and authority all lived at the point of action, with SCADA and DCS providing monitoring, not permission. Modern AI deployments inverted that: inference moved to the edge, but authority — policy enforcement, decision logging, compliance attestation — stayed in the cloud. The network changed from an enhancement into a structural dependency.
Running a model locally is not the same as governing it locally, Yates points out. When the link drops — and industrial sites routinely experience denied, disrupted, and intermittent connectivity — an AI system keeps making decisions while the governance layer that should constrain and log them is unreachable. He anchors the argument in real incidents: a 2023 incident at a BHP coal mine in Australia where an autonomous haul truck resumed movement during a communications outage, and the Cruise robotaxi case, where an autonomous vehicle made a catastrophic decision even while fully connected.
His practical prescription: governance policy enforcement, decision logging, and attestation must execute locally at the point of action, with state reconciliation when connectivity restores — the same pattern that made PLCs dependable for decades, applied to non-deterministic decision-makers. Engineers should ask vendors for partition operating specifications, document their absence as a finding, and add AI governance to design review checklists alongside cybersecurity and functional safety.
Source: Steve Yates, P.E. via NSPE PE Magazine
Read the full article here: https://www.nspe.org/career-growth/pe-magazine/issue-2-2026/the-engineer-s-duty-the-age-industrial-ai
Why it matters to Intelligent ShopFloor readers: Our deployments bring AI to real factory floors where connectivity is never guaranteed. Designing edge autonomy with local governance — what the system may decide, how it logs decisions, and how it recovers when the network returns — is what separates a demo from a production-ready system. This is a strong reference to bring into design reviews with customers.