BY IAN JONES,
COMMIT WORKS HEAD OF PRODUCT
Artificial intelligence promises transformational outcomes for mining, but without curated, connected, and contextualised data, its impact will remain theoretical rather than operational.
This image was taken in the Long Room at Trinity College, Dublin. It reminded me of the current state of Artificial Intelligence in the mining industry. AI is a technology with the potential to change the world. It is also one struggling to source the correct data to deliver its undoubted value.
For context, the Long Room is normally home to some of the most valuable works of literature, science and engineering ever written. When this photograph was taken in 2024, those works had been temporarily removed for cleaning and restoration. The statues lining the room are busts of the most famous alumni of the college. Each is accompanied by notes describing their significant contribution to their respective fields.
Data foundations matter.
So why does this scene so poignantly reflect the current state of AI in the mining industry, to me?
First, the globe positioned centrally is a reminder that, if AI is delivered correctly, the whole world will benefit. Realising this requires a common data solution that can be accessed globally, bringing together data from many systems. Much has been written about data lakes and data warehouses, but the key, for mining, is capturing broad and representative data from across operational, planning, maintenance and execution systems.
Second, while our shelves are not empty, there is a significant gap. Much of our data requires digitisation, curation, repair and contextualisation to make it fit for computer analysis. This is the domain of data architects and data scientists. In a traditional library, like the Long Room, an indexing system allows readers to locate and interpret data that points to the literature or answers they seek. Mining, by contrast, still lacks coherent data messaging solutions and, in many cases, the willingness to share data across systems or vendors. The excellent work of the Global Mining Guidelines Group (GMG) and ISA-95 working groups is starting to address this gap, but there is considerable progress to be made.
Third, data alone rarely delivers high-quality results without expert guidance and guidelines. Domain knowledge must inform how data is structured, interpreted and applied as it begins to fill the shelves. This is where the academics and experienced industry experts add significant value. Data science may be about data, but applying insights effectively is about mining, operational, and maintenance expertise. Many data projects are run by IT and Data teams. While they deliver technical solutions, they often lack operational knowledge required to curate data effectively or embed the rules, constraints and learnings that allow AI solutions to deliver meaningful outputs.
Practical steps to strengthen data foundations.
I have worked with data for many years, predating my involvement in mining by more than 10 years. My recent completion of a Graduate Certificate in Data Science did not reassure me that we are adequately preparing students for the challenges they are likely to face. In the mining industry, data complexity, context and consequence matter materially.
Based on my experience, I offer the following recommendations:
- Encourage genuine data collaboration between vendors. The richness of available data is limited only our ability to effectively integrate it across systems.
- Recognise and retain expert, industry knowledge of the experts. Too much knowledge is lost by moving engineers off projects that could unlock meaningful data insights. These experts should be engaged to develop the rules engines used to categorise and interpret data.
- As with Time Usage Models, develop standardised data categorisations for aligned to industry. These models can deliver value across multiple projects and shouldn’t need to be recreated by every organisation.
- Attribute data at a granular level. For example, don’t store only a location label, but ensure underlying information and attributes of the location can be captured and queried.
- Test assumptions rigorously using training data aligned to expected outcomes. Machine Learning may uncover some hidden gems, but poorly defined models are likely to miss the very issues you are looking intending to reveal.
It’s an exciting and generational opportunity for our industry. Let’s take time to build strong foundations, respect domain expertise, and collaborate more effectively. When we democratise the data we can make better decisions, extend machine life and, most importantly, keep our people safe.
