Just caught something interesting at the recent mining conference in Vegas. There's this investment principal, Ekin Ober, who's been doing fascinating work on how AI is actually reshaping the critical metals space. What caught my attention wasn't just the tech angle — it's how she's thinking about the intersection of innovation, sustainability and real operational challenges in mining.

So here's the thing: mining has traditionally been slow to adopt new tech, right? But Ekin Ober sees that changing rapidly with generative AI. The real bottleneck though? Getting industry stakeholders to actually understand the value proposition. As she put it, people don't need to be tech experts, but someone needs to show them how these tools work and address their concerns. Seems like even conservative markets are now dedicating serious discussion time to this, which signals the shift is real.

What's actually happening on the ground is pretty compelling. Exploration companies are using machine learning to analyze geological data and cut down exploration timelines dramatically. Think of it like this — traditional drilling uses 3000 liters of diesel per hole. If you're using computer vision to scan cores and reduce unnecessary drilling, you're talking thousands of hours of energy savings. Meanwhile, the majors like Rio Tinto, BHP and Freeport-McMoRan have deployed autonomous haul trucks and predictive maintenance systems that cut downtime and fuel consumption by 15 percent while boosting throughput. BHP's Escondida mine reportedly saved over 3 gigaliters of water and 118 gigawatt hours of energy since 2022.

Here's where Ekin Ober's perspective gets interesting: people worry about AI's energy footprint. Fair point. But consider this — one billion daily AI prompts use 340 megawatt hours of electricity, while a single mining site can consume 1000-5000 megawatt hours. So if AI helps optimize operations and reduce unnecessary energy-intensive processes like grinding (which accounts for 70 percent of mine electricity use), the technology actually reduces overall energy intensity.

Beyond operations, governments are getting in on this too. DARPA's CriticalMAAS initiative and collaborations with the US Geological Survey are using AI to automate geological map processing — cutting what used to take years down to days. The Pentagon even has an AI-driven metals forecasting program now that models supply chains and policy scenarios for critical minerals like rare earths, nickel and cobalt.

But here's what really stood out from Ekin Ober's work at Kinterra Capital: permitting. Mining projects stall for years not because of technical issues, but because nobody has bandwidth to process mountains of documents. Kinterra built a closed-loop system using large language models trained on their own criteria — permitting stages, Indigenous engagement, community sentiment. The system filters thousands of data points from filings, news and emails, then summarizes jurisdiction-specific updates directly into Teams. Real-time, actionable intelligence.

The security concern? Valid question, but Ekin Ober makes a solid point — we already trust Google, Microsoft and Apple with sensitive data daily. With legitimate tools and strong policies, it's manageable. Plus Kinterra designed their system to be traceable. You can click through to the original document and verify the quote. That transparency matters for regulators and capital providers who need confidence in decision timelines.

The broader insight here is that AI won't replace people, but it can accelerate slow government processes and get stakeholders to decision points faster. In a sector where permitting delays can cost years and billions, that's genuinely valuable. Worth watching how this evolves across the mining space.