- Essays··7 min read
23 Wells that missed an update … oops 😅
When Baker Hughes deployed Leucipa across Expand Energy's 5,782 gas wells in January 2026, the press announcement omitted 23 Marcellus wells — not from oversight, but because their Honeywell SCADA units serialised timestamps in a deprecated binary format the ingest pipeline could not parse. The piece uses this single edge case to examine the actual cost of industrial AI deployment at scale: the coverage gap that every vendor accepts and few operators audit.
oil-and-gasindustrial-ailegacy-systemsscadaRead - Essays··8 min read
What Sonepat Took Off India's Peak
HUL's Sonepat site entered the WEF Global Lighthouse Network in June 2026 with a 29 per cent energy-efficiency gain — an achievement that changes rather less about India's grid than the headline implies. Industrial electricity demand is rising at seven per cent a year, so per-site improvements are banked and spent almost simultaneously. India has roughly 1.59 crore registered manufacturing MSMEs and approximately fifteen Lighthouse sites — one per million.
indian-manufacturingindustrial-aiglobal-lighthouse-networkenergy-efficiencyRead - Essays··8 min read
The Defect You Cannot Photograph
Industrial AI vision inspection vendors cite 97–99% detection accuracy — a figure almost always trained on defects the line never made. Synthetic data can cover rare failure modes, but only after someone characterises the production environment precisely enough to generate physically plausible images and then audits the output by hand. The months nobody budgets for that work are what separate a convincing pilot from a system that holds under real production variance.
industrial-aicomputer-visionsynthetic-datamanufacturingRead - Essays··11 min read
When the Model Works and the Line Does Not
Seventy-seven per cent of AI vision pilots in manufacturing never reach full deployment — not because the detection accuracy is wrong, but because the factory floor is. This essay traces six failure modes: environments the training set never saw, integrations that ran eight times over budget, thresholds nobody will own in writing, and models that drift when procurement switches supplier. The automation thesis survives the pilot. The line does not.
ai-in-manufacturingcomputer-visionindustrial-aimodel-deploymentRead - Essays··8 min read
The Methane AI That Actually Works · The Industry That Won't Respond
UNEP's Methane Alert and Response System has processed over 1.3 million satellite observations since 2023, with AI models correctly flagging 80–85% of confirmed methane detections before expert review, helping mitigate an estimated 1.2 million tonnes. Yet 88% of MARS alerts still go unaddressed. The constraint is no longer algorithmic; it is the choice, made repeatedly by governments and operators, about which risks merit a response.
methane-emissionssatellite-monitoringclimate-aienergy-governanceRead - Essays··7 min read
The Gulf's Industrial AI Curriculum Is Missing Its Middle Rung
Across the Gulf's flagship industrial operators, public commitments to AI reskilling have assembled a recognisable two-track system: research master's programmes at KAUST and MBZUAI at the top, SDAIA and ADNOC bootcamps at the bottom. The publicly disclosed plan documents have no middle rung, no plant-embedded practitioner who can translate an anomaly-detection alert into a shutdown call at a DCS console at 3am.
gulf-workforceindustrial-aiskills-gapai-educationRead - Essays··7 min read
The Cleanest Air in India Costs the Most Electricity
India's largest generic pharmaceutical manufacturers now procure seventy to 85% of their electricity from renewables, a figure that flatters the underlying problem. Pharmaceutical cleanrooms account for fifty to 80% of a sterile plant's total energy and run continuously at setpoints regulators inspect without warning; changing the grid mix does not move the meter. Model predictive control of chilled-water loops is where a ten-to-20% reduction is now measurable.
pharmaceutical-manufacturingrenewable-energyhvacindiaRead - Essays··8 min read
Sixty Picks an Hour, and the 88% That Never Graduate
Siemens' Erlangen humanoid moved sixty totes an hour at better than 90% accuracy. Across enterprise, 88% of AI pilots never reach production at all. The marquee deployments are three dozen Tier-1 buyers reading from one address book; everyone else is watching pilots die in the gap between a vendor demo and a system that still runs on a Tuesday in August.
ai-in-manufacturinghumanoid-robotspilot-to-productionindustrial-aiRead - Essays··7 min read
What's Actually on the Floor in Erlangen, Leipzig, and Spartanburg
Three press releases in three months told the industrial-AI story more clearly than any analyst report. Erlangen, Leipzig, and Spartanburg have settled whether the systems function. The open question is whether the payback survives an audit.
industrial-aihumanoid-robotsdigital-twinopc-uaRead - Essays··7 min read
Hannover Messe 2026 delivered an industrial AI arms race · copilot versus agent
The fork between copilot and agent is the right framing for the next two years of industrial AI. The technology is real. The governance architecture to match it is not yet built.
industrial-aiagentic-manufacturingdata-governancehannover-messe-2026Read - Essays··8 min read
Erlangen, Spartanburg, and the Quiet Half of Industrial AI
The cognitive layer of industrial AI (copilots, agentic optimisation, the operating-system metaphor) is real and mostly in pilot. The physical layer of vision, manipulation and motion is the part actually booking cycle time on the factory floor. That asymmetry is the story.
industrial-aipilot-to-productionembedded-visionhumanoid-roboticsRead