The progression matters. Connected OEE data comes first, then data science finds correlations, then advanced process control acts on them. Skipping to sophisticated AI without that foundation means the models have nothing reliable to work with.
@fogoros
-
AI process control in CPG plants needs connected data infrastructure
By
–
Advanced process control has performed this function in CPG plants for years, described as the first generation of AI. More sophisticated AI models follow, but they depend on the same connected data infrastructure that most CPG plants have not yet built. pic.twitter.com/nf5RtN60da
— Lucian Fogoros (@fogoros) 7 octobre 2026Advanced process control has performed this function in CPG plants for years, described as the first generation of AI. More sophisticated AI models follow, but they depend on the same connected data infrastructure that most CPG plants have not yet built.
-
AI-assisted root cause diagnosis needs full-resolution historical data
By
–
AI-assisted root cause diagnosis produces solutions that last rather than band-aid fixes, but only when the historical data infrastructure preserves the full-resolution records the model needs to compare against.
-
ML optimization in dosing system reveals hidden process relationships
By
–
The dosing system example shows ML optimization in practice: humidity, ambient conditions, and cross-process data all influenced outcomes that engineers had previously tuned manually without knowing those relationships existed.
-
AI risk register for OT environments and ownership
By
–
Ian Bramson of Black and Veatch proposed a third angle: locate your risk register for AI in OT, identify who owns each risk listed, and determine the priority order. A well-developed risk register can go all the way to the board level.
-
Data layers must feed AI models across the business
By
–
The data and contextualization layers must be addressed first, because operational data trapped in individual sensors, PLCs, or controllers cannot feed AI models across the business.
-
Capturing tacit knowledge for industrial AI agents at commissioning
By
–
Commissioning is the moment to capture tacit knowledge for industrial AI, says Cybus CEO Peter Sorowka, because the person wiring up the machine knows what it is and does. Capture once then, and the agent reads the model instead of asking people repeatedly. Partner content with… pic.twitter.com/Q6swxpE56p
— Lucian Fogoros (@fogoros) 5 octobre 2026Commissioning is the moment to capture tacit knowledge for industrial AI, says Cybus CEO Peter Sorowka, because the person wiring up the machine knows what it is and does. Capture once then, and the agent reads the model instead of asking people repeatedly. Partner content with
-
Predictive maintenance AI lacks fault identification and action guidance
By
–
Predictive maintenance outputs a probability, a risk score, a trend alert, a threshold breach. But it does not identify which component, what fault mode, what evidence supports the conclusion, or what action to take by when. Partner content with @AI4ProdOutcomes. #infinite_ai pic.twitter.com/a7gUVQG7WM
— Lucian Fogoros (@fogoros) 1 octobre 2026Predictive maintenance outputs a probability, a risk score, a trend alert, a threshold breach. But it does not identify which component, what fault mode, what evidence supports the conclusion, or what action to take by when. Partner content with @AI4ProdOutcomes
. #infinite_ai -
Context-trained decision-making for robots is scoped, not open-ended
By
–
Context-trained decision-making for robots is a very different architecture than a general-purpose model. The clear context framing matters: it means the deployment is scoped, not open-ended.
-
TCS’s Besprosvan: AI supports factory workers, not replaces, via hazard reduction ROI
By
–
Sol Besprosvan from TCS framed it as AI supporting professionals on factory floors, not replacing them. The use case anchor is hazard reduction, which is where the ROI conversation is easiest to win internally.