Source: Kay Sever | July 29, 2026
So far in 2026, I have shared what I have learned about AI capabilities (which change continually), AI components, AI metrics, and the complexities of linking AI to existing IT systems. I have also discussed corporate risks that may not be discussed by AI developers (accuracy, depth of knowledge, resources needed, AI integration timelines, what data to share, how to protect its confidentiality, etc.)
The last half of this year we are covering examples of applications where AI might be used and talk about barriers and benefits as I understand them today. Last month I scratched the surface on analyzing production data… actual data that is currently collected on production equipment and processes, plus other data that must exist and be shared with AI to yield meaningful outputs that could be used for decision-making.
This month we are going to crack open the lid on using AI to identify and quantify problems in the workplace (Part 1). Because this topic is so broad, I am starting with production problems in general. To understand what AI capabilities are required, let’s look at how people identify a production problem:
Unexpected variation is one of the first indicators of a possible problem… process variation that exceeds defined min/max limits OR variation from a standard (construction/design specifications, cost, volume, input quality, output quality, density, temperature or other process characteristics.) If people are watching a process perform real-time and are expecting a certain consistency or result but do not see the expected result, they know there is a problem somewhere based on what is defined as acceptable/normal operations.
Once a problem has been identified, it is ranked/prioritized so a company can determine which problems to tackle 1st, 2nd and 3rd. Quantifying the loss caused by a problem is the most effective way to rank problems because stopping the losses means more profit. If the problem has been reoccurring over time, the loss calculation must include losses for each occurrence over time. What appears to be a 10-minute problem might be generating a million dollar (unreported) loss depending on the frequency of occurrence. Such production problems do exist and are ignored or overlooked by management because a single event is deemed to be unimportant to fix. (Twenty years ago I published an article titled “The 10-Minute Million Dollar Problem” where I described an actual problem I had helped identify and solve… one that had been overlooked as minor for several years. Millions of dollars in unknown losses had occurred over time without management’s knowledge.)
There are different methods for quantifying the cost of a problem. Lost production and excess costs are two parts to the calculation. At the end of the day, how people and companies measure SUCCESS determines the size of the losses AND the corrective action taken to fix the problem.
- If your goal is achieving budget, you measure success by the size of the budget variance… how close you were to achieving the plan for the year. It is important to be aware that budgets contain hidden losses for unsolved problems. If you use budget variances to estimate a loss cause by a production problem, you are likely to underestimate its cost significantly.
- If your goal is achieving optimum (best possible), you would determine a loss based on the difference between actual and what was possible to achieve. This loss is real because you would already have the equipment and people in place to achieve more but failed to do so. You would have values for “best” levels of performance based on a standard process for determining them. These losses would represent your upside that is there for the taking without spending capital on new equipment or expansions to get it. If you make a “free” change to stop the losses, profit goes up by that amount. This focus makes budget easier to achieve.
Can AI Help Identify and Prioritize Problems?
With all of the above said, let’s go back to our original question. We know that people can observe and measure variation. We also know that people can calculate financial losses caused by problems, then prioritize those problems for corrective action. What would AI need to access to perform these functions? I have given this question some thought and have shared my opinions about the answer below:
- ACTUAL DATA: AI would have to have access to your production history files. Since much production data is captured in decentralized systems across a mining site, multiple interfaces would probably have to be built. For example, all mobile production equipment data might flow into one mine productivity system, but the maintenance system likely stands alone. Crushing/mill data would be in a separate system. Downstream plants for smelting, refining, etc. would also likely be captured in stand-alone systems of varying sophistication. Since problem identification depends on variation, detailed levels of historical data would be required by AI to identify shortfalls or excesses over time based on limits that would also have to be defined for AI’s use. Process changes over time would impact process capability over time. AI would have to know when process capability was changed/expanded to identify losses correctly. To value the problems identified, AI would have to know how to assign values to losses, have access to detailed financial records for costs, and know how to separate excess costs from true variable costs linked to the production process.
- BUDGETS: If you wanted AI to analyze budget variances, it would have to have access to your budget files and would have to be able to perform the same kind of analyses as your financial analysts do now. AI would have to be able to determine which variances to focus on, which would vary month to month.
- OPTIMIZATION: If you wanted AI to identify optimization shortfalls, “best” values often kept in spreadsheets would have to exist and be accessible by AI across the value stream. Calculations performed in spreadsheets would also have to be defined and accessible by AI.
- TESTING: If interfaces are built to link AI to the many systems across a mine site, significant resources would have to be dedicated to testing each interface to ensure a very high percent of the results returned by AI could be trusted and used for decision-making.
My Conclusion: The cost and time required to build and test the many AI interfaces required for identifying and quantifying production problems may far exceed initial estimates. People that know production processes understand the problems they encounter and can easily calculate losses linked to those problems. I believe the work required to turn these tasks over to AI would not be worth the investment.
Next month: Can AI determine corrective action to fix problems? (Pa
Kay Sever is an Expert on Achieving “Best Possible” Results. Kay helps executive and management teams tap their hidden profit potential and reach their optimization goals. Kay has developed a LIVESTREAM management training/coaching system for Optimization Management called MiningOpportunity – NO TRAVEL REQUIRED. See MiningOpportunity.com for her contact information and training information.Optimization/AI Integration – AI Scope: An Overview
To comment on this story or for additional details click on related button above.
- About Us
Kay has worked side by side with corporate and production sites in a management/leadership/consulting role for 35+ years. She helps management teams improve performance, profit, culture and change, but does it in a way that connects people and the corporate culture to their hidden potential. Kay helps companies move “beyond improvement” to a state of “sustained optimization”. With her guidance and the MiningOpportunity system, management teams can measure the losses caused by weaknesses in their current culture, shift to a Loss Reduction Culture to reduce the losses, and “manage” the gains from the new culture as a second income stream.
