Ekin Ober on Why AI Could Be Mining’s Most Valuable Tool Yet

How AI is being deployed

In addition to benefiting project planning through better modeling and digital twin, AI is making mining more efficient, safe and environmentally responsible.

In exploration, startups like KoBold use machine learning to analyze geological data, drastically cutting the time and cost of identifying potential lithium, copper, nickel and cobalt deposits

Operationally, majors such as Rio Tinto (ASX:RIO,NYSE:RIO,LSE:RIO), BHP (ASX:BHP,NYSE:BHP,LSE:BHP) and Freeport-McMoRan (NYSE:FCX), deploy AI-powered autonomous haul trucks, drills and predictive maintenance systems that have slashed downtime and fuel use by up to 15 percent, while boosting throughput by 10 to 15 percent.

On the environmental front, AI tools optimize water management, monitor air quality and reduce waste, BHP’s Escondida mine reportedly saved over 3 gigaliters of water and 118 gigawatt hours of energy since 2022.

While AI isn't without its own controversy, usually arising from its energy consumption, Ober explained that AI integration can help reduce a mining site's overall energy intensity.

It is estimated that one billion daily AI prompts utilize 340 megawatt hours of electricity each day, while a mining site can use upwards of 1000 - 5000 megawatt hours. According to data from Natural Resources Canada, global mining operations consume 3 percent - 6 percent of the world's electricity.

Together, AI can help the mining sector better target deposits and reduce the amount of energy deployed.

“Drill holes (alone) use 3000 liters of diesel. And when you look at grinding, grinding ore is 70 percent of the mine’s electricity (consumption),” said Ober.

She added: So if you're using the technology for scans, you're able to use computer vision and scan a core, or look at the geography to reduce the number of drills, or the grinding exercise that you're going through, then it can actually save 1000s of hours of energy, conserving more than it consumes.”

From policy bottlenecks to permit approvals

This efficiency has made AI data sets appealing to governments as well. Through initiatives like DARPA’s CriticalMAAS and a collaboration with the US Geological Survey, AI models can now transform geologic map processing — from years to mere days — by automating georeferencing and mineral feature extraction.

These tools help rapidly assess hundreds of critical minerals across vast regions, accelerating decision-making and reducing exploration risk.

Meanwhile, the Pentagon’s AI-driven metals forecasting program, now managed by the Critical Minerals Forum, models supply, pricing and policy scenarios to bolster US sourcing strategies — especially for rare earths, nickel and cobalt.

For Ober, AI can also be integral to the often extended permitting process, while also implementing ESG goals and best practices. She explained that at Kinterra, AI is already playing a key role in streamlining permitting assessments, one of the most complex hurdles in mine development.

The firm has built a closed-loop system using large language models layered with its own criteria and values, including permitting stages, Indigenous engagement and community sentiment. The tool filters thousands of data points — from state filings to news releases and emails — extracting only what’s relevant.

Jurisdiction-specific updates are then summarized and delivered directly into Microsoft Teams, offering a real-time, digestible overview of key permitting signals.

“We need the company and the community to be engaged,” she said. “We take a very proactive approach. We engage very early on.”

Industry wide Ober sees AI improving the efficiency and transparency of mining permitting.

“One of the biggest concerns we hear is around security,” said Ober. “But we already trust companies like Google, Microsoft and Apple with sensitive data every day. If you’re using legitimate tools with strong policies in place, it’s manageable.”

Ober believes AI’s biggest value lies in its ability to accelerate slow, document-heavy government processes.

“Permitting can stall a project for years — not because of technical issues, but because no one has time to read the documents,” she said. “That’s where AI can help. Large language models can extract key information, layer in governance or environmental criteria and summarize it in a way that’s actionable.”

To address the risk of accuracy, Kinterra has designed its systems to generate traceable outputs.

“You can click a link and go straight to the original document and quote,” she explained, adding that this level of transparency is crucial for regulators and investors alike.

“It’s hard to commit capital when you don’t know if or when a permit will be granted,” she said. “AI won’t replace people, but it can get us to decision points faster — something the entire sector needs.”