Eric Decker
Artificial intelligence is getting plenty of attention in mining. Vendors promise transformation, conference agendas are full of it and leaders feel pressure to act. But for most of the industry, it is still early and much remains unproven. That is not a reason to sit on the sidelines. It is a reason to start carefully, measure what works and learn as you go. We are on that journey ourselves and a few early lessons may help others decide where to begin.
Guide It, Don’t Dam It
Our first instinct was caution, so we blocked public generative AI tools to protect company information. The restriction did not stop adoption. Employees were already using these tools in their personal lives, understood their value and found ways around the block. That experience showed us two things, the demand was real and generative AI could be a genuine productivity multiplier if we gave employees a secure, approved way to use it.
Anyone who has managed water on a mine site understands what came next. You do not manage water by simply putting a wall in front of it, you channel it where it can move safely. We took the same approach with AI. We approved a set of tools with the security and data protections our business requires, directed employees toward them and supported that access with a clear usage policy and training on what can and cannot be shared.
Guardrails matter, but the goal is to be cautious without being paranoid. Like a good safety program, governance should help people work confidently, not simply tell them no.
Let Champions Bring You the Problems
In every organization there are people already leaning into AI. Find them early. They will promote practical use among their peers and spot opportunities leadership might miss. Our first operational use case came this way. After one of our AI discovery sessions, a maintenance manager approached us with an idea, use maintenance history and oil analysis data to predict component failures in our haul truck fleet. He understood the problem; knew the data existed and was motivated to see it work.
“Mining has always rewarded steady, disciplinedimprovement. AI should be approached the same way,start with a real problem, learn from the result and build capability over time.”
That is the kind of first project to look for, a real operational problem, an engaged business owner and enough usable data to test whether the idea has value. The first project does not need perfect data, but it does need a result the operation can recognize and measure. Partners can bring specialized skills quickly but build an internal team alongside them so your organization learns to understand, question and eventually own the work.
What Our First Pilot Taught Us
Early analysis produced results that looked promising, but it is too soon to declare success. We are only now finishing the rollout across the full fleet at our initial test site. So far, the system has flagged maintenance on components we already knew needed attention rather than uncovering a new issue. The major catch that would deliver the largest savings has not happened yet. Proving that value will require more time and operating data.
What we did not expect was everything else. We had data everywhere, but very little of it was driving decisions because the systems holding it were difficult to use. It was a case of “water, water everywhere, but not a drop to drink.” Simply focusing on one problem put far more usable information in front of our maintenance teams. We now catch standard work that might otherwise be missed. Our people have become more curious and data hungry and more decisions are made on evidence rather than gut feel. That shift may prove more valuable than any single prediction. As data-driven decision making matures, we will rely less on the decades of experience that only a few senior supervisors carry today, an important advantage as our industry’s workforce changes. Over time, we expect machine learning to play a larger role in those decisions. So do not be disappointed if your early wins look more like better data visibility than artificial intelligence. That is real progress and it builds the foundation AI depends on.
A Journey, not a Purchase
Be careful not to overpromise. Your first use case may not knock it out of the park. If it shows improvement, you are headed in the right direction. Meaningful results usually come from many cumulative gains rather than a single breakthrough and as your capability grows, you will see opportunities that were not visible at the start. Mining has always rewarded steady, disciplined improvement. AI should be approached the same way, start with a real problem, learn from the result and build capability over time. Used well, it does not replace experienced people. It gives them better information for the decisions they make every day.
