{"id":6381,"date":"2026-02-06T09:30:12","date_gmt":"2026-02-06T09:30:12","guid":{"rendered":"https:\/\/newtonimpact.com\/?post_type=content_library&p=6381"},"modified":"2026-02-06T10:26:45","modified_gmt":"2026-02-06T10:26:45","slug":"industrialising-ai-in-asset-intensive-industries-from-pilots-to-operational-advantage","status":"publish","type":"content_library","link":"https:\/\/newtonimpact.com\/en-us\/content-library\/blog\/industrialising-ai-in-asset-intensive-industries-from-pilots-to-operational-advantage\/","title":{"rendered":"Industrialising AI in asset-intensive industries: from ‘pilots’ to operational advantage"},"content":{"rendered":"\n
\n
\n
\n \n
\n Take Me Back<\/a>\n\n

Industrialising AI in asset-intensive industries: from ‘pilots’ to operational advantage<\/h1>\n

By Craig Hoggett<\/p>\n\n

\n \n \n

Blog<\/p>\n <\/span>\n \n

5 min<\/p>\n <\/span>\n \n

6th February 2026<\/p>\n <\/div>\n<\/div>\n\n <\/div>\n <\/div>\n <\/section>\n\n \n\n\n

\n
\n
\n \n
\n
\n \"\"\n <\/div>\n <\/div>\n\n <\/div>\n <\/div>\n <\/section>\n\n \n\n\n
\n
\n
\n \n
\n \n
\n

Executive teams across\u00a0Energy, Utilities,\u00a0Transport\u00a0and the\u00a0extractive industries\u00a0have moved beyond the \u201cshould we do AI?\u201d debate.\u00a0\u00a0The strategic question now is more practical:<\/p>\n

How do we industrialise AI safely and profitably,\u00a0without creating another graveyard of pilots?<\/span><\/b>\u00a0<\/span><\/p>\n

The evidence is clear: AI can deliver meaningful operational and financial outcomes at scale, but only when it is treated as\u00a0<\/span>an operating discipline<\/span><\/b>,\u00a0<\/span>not<\/span><\/i>\u00a0a technology experiment.\u00a0<\/span>\u00a0<\/span><\/p>\n

This\u00a0article looks at\u00a0what\u2019s working,\u00a0what\u2019s\u00a0holding organisations back, and\u00a0how Newton\u00a0partners\u00a0with\u00a0organisations\u00a0to\u00a0unlock value\u00a0quickly, safely, and in a way\u00a0that\u00a0scales.<\/span>\u00a0<\/span><\/p>\n

What the evidence says is working now<\/span><\/b>\u00a0<\/span><\/p>\n

1) Industrial value is real:\u00a0when AI is anchored in operations<\/span><\/b>\u00a0<\/span><\/p>\n

One of the clearest demonstrations of scale comes from a southern US utility that deployed\u00a0<\/span>400+ AI models across 67 generation units<\/span><\/b>, reporting\u00a0<\/span>$60m in annual savings<\/span><\/b>\u00a0and a reduction of\u00a0<\/span>1.6m tonnes of CO\u2082<\/span><\/b>.\u00a0\u00a0The headline matters less than the pattern: value came from embedding AI into operational routines, not from isolated dashboards.\u00a0<\/span>\u00a0<\/span><\/p>\n

The Newton view:<\/span><\/b>\u00a0The organisations\u00a0that genuinely\u00a0benefit\u00a0from\u00a0AI,\u00a0use\u00a0it in\u00a0a similar way\u00a0to\u00a0reliability engineering or continuous improvement, i.e.\u00a0repeatable methods, clear governance, and sustained adoption.<\/span>\u00a0<\/span><\/p>\n

2) Predictive maintenance\u00a0remains\u00a0the \u201chighest confidence\u201d starting point<\/span><\/b>\u00a0<\/span><\/p>\n

Across industries, predictive maintenance shows some of the most consistent economic returns,\u00a0particularly for\u00a0<\/span>rotating equipment<\/span><\/b>\u00a0(turbines, compressors, pumps, fans).\u00a0\u00a0Our research over the past two years\u00a0report benefits\u00a0<\/span>up to ~30% maintenance cost reduction<\/span><\/b>\u00a0and\u00a0<\/span>~20% availability improvement<\/span><\/b>\u00a0when implemented well.\u00a0<\/span>\u00a0<\/span><\/p>\n

What distinguishes success:<\/span><\/b>\u00a0<\/span><\/p>\n

Start with critical assets that have sufficient sensor coverage and history.<\/span><\/p>\n

Integrate outputs into the CMMS\/EAM workflow (alerts\u00a0<\/span>\u2192<\/span>\u00a0work orders\u00a0<\/span>\u2192<\/span> feedback loop).<\/span><\/p>\n

Use reliability engineers to validate early results and manage false positives.<\/span><\/p>\n

3) GenAI\u2019s early ‘killer app’ is knowledge retrieval – not autonomy<\/span><\/b>\u00a0<\/span><\/p>\n

In asset-intensive environments, productivity is routinely lost to searching for the right information: SOPs, manuals, standards, incident reports, P&IDs.\u00a0\u00a0Retrieval-Augmented Generation (RAG) systems in industrial contexts have demonstrated strong retrieval performance for technical queries (e.g.\u00a0<\/span>88% mean reciprocal rank<\/span><\/b>) and for internal regulatory content (e.g. <\/span>91.62% recall<\/span><\/b>).\u00a0<\/span>\u00a0<\/span><\/p>\n

The Newton view:<\/span><\/b>\u00a0In 2026, the pragmatic path is to deploy GenAI where it reduces friction and improves consistency,\u00a0while keeping humans firmly accountable for decisions, especially in safety-critical contexts.\u00a0<\/span>\u00a0<\/span><\/p>\n

4) Capital delivery is being reshaped,\u00a0if\u00a0(and only if)\u00a0data and governance are ready<\/span><\/b>\u00a0<\/span><\/p>\n

AI tools for schedule risk analysis are now trained on\u00a0substantial\u00a0historical datasets (e.g.,\u00a0<\/span>750,000 schedules<\/span><\/b>\u00a0representing\u00a0<\/span>>$2T<\/span><\/b>\u00a0in construction spend).\u00a0\u00a0This creates a credible foundation for early\u00a0detection of schedule risk patterns,\u00a0but value\u00a0is very dependent\u00a0on data quality, process integration, and decision rights.\u00a0<\/span>\u00a0<\/span><\/p>\n

The Newton view:<\/span><\/b> Capital analytics can be a major lever, but organisations should validate ROI in context and avoid ‘vendor case study optimism’ without evidence from their own portfolio. <\/span>\u00a0<\/span><\/p>\n <\/div>\n<\/div>\n\n <\/div>\n <\/div>\n <\/section>\n\n \n\n\n

\n
\n
\n \n
\n
\n

“Newton\u2019s contribution is not ‘more AI for the sake of AI’. It is unlocking complexity for meaningful impact and helping clients industrialise improvements that already fit their reality, technically, operationally, and culturally.”<\/p>\n \n Craig Hoggett <\/cite>\n <\/blockquote>\n<\/div>\n\n <\/div>\n <\/div>\n <\/section>\n\n \n\n\n

\n
\n
\n \n
\n \n
\n

The biggest constraint\u00a0isn\u2019t\u00a0algorithms, it\u00a0is\u00a0<\/span><\/b>adoption<\/span><\/i><\/b>.<\/span><\/b>\u00a0<\/span><\/p>\n

The human factor is the determining factor in many cases:<\/span>\u00a0<\/span><\/p>\n

~85% of AI projects fail <\/span><\/b>to deliver expected business value, often due to inadequate planning, training, and change management.<\/span><\/p>\n

~80% of workers report receiving no GenAI training<\/span><\/b> from their organisations.<\/span><\/p>\n

Meanwhile, many asset-intensive organisations\u00a0operate\u00a0in brownfield reality: imperfect historic\u00a0data, legacy OT constraints, and cybersecurity architectures not designed for modern connectivity.\u00a0<\/span>\u00a0<\/span><\/p>\n

The Newton view:<\/span><\/b>\u00a0The difference between success and failure is usually not model selection. It is:<\/span>\u00a0<\/span><\/p>\n