AI Hype X Reality - What's Real, What's Working, and What's Next

Only 20 to 25% of manufacturers have moved into smart manufacturing, and most are still years away from what anyone would call AI-enabled operations. This AI Manufacturing Day 2026 panel separated the real from the hype, examining what is actually working on factory floors today and why scaling remains so difficult. Hamish Mackenzie (IIoT World) moderated a conversation with Andrew Scheuermann (Arch Systems), Gary Tillery (Skkynet), Dave Morse (Delta Electronics), and Pugal Janakiraman (Snowflake). The panel covered everything from the protocol gap between OT and IT to why change management, not technology, determines whether AI scales. Gary Tillery explained that much of what is being sold as agentic AI is just better visualization with a generative UI layer on top. Real agentic AI perceives, reasons, and acts in a closed loop. Any write-back into the OT environment must be authenticated, audited, and architecturally constrained so a compromised agent does not become an attack vector. He identified two foundational problems the industry skipped: the protocol reality, where AI expects clean JSON and MQTT but the OT floor speaks OPC DA, Modbus, and proprietary historian formats, and security. Andrew Scheuermann defined what is real versus hype. Real is specific use cases paired with the subset of data that is ready. Not real is the claim of a fully unified data model across the entire factory. He identified three common investment mistakes: doing nothing, over-investing in connectivity for connectivity's sake, and building a pure chatbot strategy without embedded workflows. His team observed that manufacturers now want 4 to 5 proven use cases before committing to the change management effort required to scale, because the ROI must justify that organizational cost. Dave Morse shared a current example: a full assembly line designed and simulated on a digital twin, manufactured from it, and now being installed in Texas. It works, but it is not connected to the ERP. Raw materials cannot get to the line, and finished product cannot get off it into the warehouse. The end-to-end integration is missing. He estimated 30% or more of manufacturers still have no automation at all. Pugal Janakiraman noted that 60 to 70% of root cause analysis time is spent on analytics alone, representing a clear opportunity for AI. But scaling across plants is blocked by the reality that no two facilities share the same IT and OT footprint, even within the same company, especially across acquisitions in different regions. This session was sponsored by Arch Systems and Skkynet. Editorially independent. Subscribe to our newsletter: https://www.iiot-world.com/subscribe/ IIoT World: https://www.iiot-world.com LinkedIn:   / iiot-world   https://x.com/iiot_world #AIManufacturingDay #ManufacturingAI #AIHypeVsReality