DATA & AI · INDUSTRIAL AI SUMMIT

The layer nobody built: why factory data never reaches AI

Ten to twenty years of sensors and connectivity, and the data still doesn’t land anywhere an AI model can use it. At the Industrial AI Summit 2026, Jonathan Alexander (Albemarle), Peter Sorowka (Cybus) and Kudzai Manditereza (HiveMQ) mapped where it gets stuck. Collection is solved. Historisation is widespread. Contextualisation is simply missing — the equipment relationships and operating conditions that make a reading mean something still live in the heads of your operators, not inside any system. Their second point lands just as hard: point-to-point, application-centric integration is why so many plants never get past the pilot. Every new use case pays for its own plumbing.

OT SECURITY · NEW RESEARCH

68% can’t see their own OT assets in real time

Palo Alto Networks’ 2026 State of Critical Infrastructure Cybersecurity report surveyed more than 1,600 security leaders across 11 countries and five sectors, and it is uncomfortable reading. 60% had a significant breach in the past year — half of those reporting physical safety concerns as a result. 68% lack complete, real-time visibility of everything connected to their OT networks, and 42% name legacy, unpatchable OT assets as their single biggest risk. Then the timing gap: 29% of 2026 CVEs were exploited inside 24 hours, against an industry average of 55 days to deploy a patch. Tool sprawl is part of the problem — seven disparate security systems on average — and 74% still have not integrated IT and OT security operations.

ROBOTICS · PHYSICAL AI

A 99% robot is not a 99% automated job

A useful antidote to humanoid demo season, from Nicolas Sauvage of TDK Ventures. Take a workflow of 100 sequential physical actions. At 99% success per action, the odds of getting through all 100 first time are 36.6%. At 99.9% it is 90.5%; at 99.99% it reaches 99.0%. Small reliability gaps compound brutally across a real job, which is why a robot that looks flawless on stage can still be nowhere near ready to work unattended. His argument: dexterity is a closed-loop system — vision, position, contact, force, grasp, movement, recovery — not a hand with more fingers. The customer does not buy degrees of freedom. They buy useful work.