COLORADO SPRINGS, Colo. – Fifteen of 19 major mining companies tracked by McKinsey reported financial gains from artificial intelligence in the third quarter through Sept. 11, up from four in the previous quarter as the technology began moving beyond pilot projects.
AI could lift miners’ earnings before interest, taxes, depreciation and amortization by 10%-15% by raising production about 5% and cutting direct C1 cash costs about 10%, The New York-based consultancy estimates. Processing leads, with 10 of the 19 companies reporting a bottom-line impact from AI in their plants.
“A pile of AI initiatives is not a transformation,” Ferran Pujol, a Santiago-based McKinsey partner who leads its Global Energy & Materials practice in Latin America, told Mining Forum Americas in Colorado Springs on Monday. “The test is whether the domain runs differently, not if we have better tools.”
The jump suggests miners may finally be converting years of experiments into operating gains. But McKinsey labels the tally “realized value claims” and doesn’t identify the miners or disclose the dollar value of the gains.
Pujol called the incomplete-quarter figure a first data point that still needs confirmation. The test now is whether companies can sustain the gains, separate them from commodity prices and normal operating changes, and repeat them across mines.
Plant gains
To assess how far mining AI applications had progressed, McKinsey divided them into four evidence categories. It reserves its highest evidence rating, “proven in the profit and loss statement,,” for an AI tool in production with a financial figure disclosed in results or a regulatory filing.
The next level, “proven,” required a company to have deployed the tool and quantified its effect. Emerging tools had met only one of those tests, while aspirational uses hadn’t entered production.
Using AI to adjust mill and flotation settings was the only processing application to reach McKinsey’s top evidence category. McKinsey rated computer vision and ore-blending and stockpile optimization as proven, while leaching and water-and-reagent optimization remained emerging.
Pujol said plants use models to track ore characteristics from the block model to the mill, choose the best blend and adjust operating settings. The most advanced companies are connecting the models directly to plant controls, shifting operators from setting variables manually to supervising and improving the models.
Practice perfect
Maintenance also scored strongly. Planning and shutdown optimization reached the top category, while predictive maintenance for mine fleets and plants ranked as proven. Despite the attention predictive maintenance receives, Pujol said most of the value is coming from planning and shutdown optimization.
Ravikanth “Ravi” Malladi, senior adviser for data and AI at Freeport-McMoRan (NYSE: FCX), said mines are gaining by servicing equipment according to its condition instead of fixed schedules or waiting for failures. But poor basic practices can defeat the technology.
“You can take an oil sample,” Malladi said. “If the oil sample doesn’t make it to the lab and the data doesn’t get into your system, which is your foundational system of records, I’m afraid AI is not able to move the needle very much.”
Evidence thins beyond the plant. McKinsey rated digital twins of mines and plants as proven but classified the other eight mine-planning and operating applications as emerging or aspirational. They included mine scheduling, drill-and-blast design, dispatch, collision avoidance, electric-fleet charging and mixed-fleet autonomy. All three exploration applications – targeting, automated core logging and grade control – remained emerging.
Akilan Kapilan, a Microsoft (Nasdaq: MSFT) director and industry adviser for energy and resources, said models should recommend options rather than make unchecked operating decisions. Metallurgists and other specialists must set limits and retain control because models can produce incorrect answers when they aren’t grounded in reliable data.
Ivanhoe Mines (TSX: IVN; US-OTC: IVPAF) founder and executive co-chairman Robert Friedland widened the discussion to AI’s effect on mineral demand. The race to develop AI would accelerate needs for electricity, copper and critical minerals regardless of how quickly miners adopt the technology themselves, he said in a separate presentation.
“I’m not sure what AI will do for the average American,” he said. “But I can assure you AI’s moving forward because it’s a military technology.”
Scaling test
McKinsey’s broader global survey of 1,719 respondents shows why the mining jump matters. Across industries, 89% said their organizations regularly used AI in at least one business function this year, but only 44% reported scaling it across the enterprise.
Just 7% of energy and materials respondents had scaled AI agents in manufacturing and McKinsey senior partner Richard Sellschop said mining probably trailed even that figure because stronger uptake in the power sector raised the group average.
Pujol said companies that move beyond pilots tend to follow a “70-70-70” staffing rule: they keep 70% of their technology talent in-house, make 70% of the team builders rather than coordinators or managers and draw 70% from senior ranks.
They also connect technology staff directly with operating teams and give one executive control of the budget and financial result. In a hybrid model described by Pujol, central teams provide data systems, platforms and reusable tools while each mine owns the value and budget because its orebody and operating conditions differ.
The chief executive should appoint that owner, while the chief operating officer spreads successful tools to other sites and the chief financial officer separates AI’s contribution from other business changes, McKinsey said. Malladi urged companies to embed technology teams with workers at mine sites rather than manage deployment from a corporate office.
For boards and senior executives, Malladi said adopting AI is no longer the question. They want to know: “When does the cash register ring?”
Comments