Everyone is being told to have an AI plan. Few leaders have been told where to start. If you run a bulk materials operation and measure with Stockpile Reports®, the answer is closer than you think: you started years ago.
Somewhere in the last year, "what are we doing about AI" stopped being an IT question and became a board question. Most leadership teams answer it by reaching for something with AI in the name: a rollout plan, a model, a pilot. The harder part is the numbers underneath, and for your stockpiles, an asset that is hard to count and one you already carry on the balance sheet, that is the part you have been working on all along. The mandate is not asking you to begin. It is asking you to notice what you already have, and to finish the job.
The model is not the hard part
In February 2025, Gartner predicted that through 2026, organizations will abandon 60 percent of AI projects that are not supported by AI-ready data. In the survey behind that prediction, 63 percent of organizations said they either did not have the right data management practices for AI or were not sure they did.
The model gets the attention. The numbers underneath it decide whether the project survives. For stockpile inventory, that comes down to three questions. Is the number current? Can the people and tools that need it reach it? Can anyone stand behind it? Those are not three separate initiatives. They are one discipline, in an order that matters, because each step is what makes the next one worth doing.
What you are already running
Stockpile Reports is the Predictive Physical AI that turns imagery from a phone, drone, or installed camera into an independent, verified report: volume and tonnage with coverage confidence, high, medium or low. It is powered by EveryPoint®, and it rests on computer vision since 2011 and machine learning for over a decade. It is the authoring layer for bulk materials: the moment physical reality becomes data, before anyone opens software to write it down.
It is not a pilot you are still evaluating. Customers have been measuring piles with Stockpile Reports since 2012. If you want the longer version of why this kind of AI matters more in a yard than the kind in the headlines, it is in Physical AI 101.
Step one: measure often enough that the number is current
Your own crew can capture the measurement. No specialist has to be hired or certified. A scan with SR Measure® on an iPhone, your own drone, or an installed camera already watching the yard: use whichever fits the pile. If you would rather not fly a drone yourself, Full-Service Flight takes that on for you. For a pile you can scan with a phone, there is nobody to book at all. An aircraft is not the problem. Having to find someone to fly it is.
This step comes first for a reason. Any answer about a pile is only as current as its latest measurement. If a pile was last measured in the spring, the answer describes the spring. As we wrote in August, the real risk on a site is not runaway AI, it is stale data. A stale measurement is not only an operational lag. What your book says and what's on the ground are different numbers. How often you measure sets how current every answer can be.
Step two: put the number where the work happens
A report is only as useful as the number of people who remember to open it. Connect Stockpile Reports to the AI assistant your team already uses, and anyone on the account can ask, in plain language, for the latest measurements at the sites they have access to, and get the recorded numbers back. The Stockpile Reports MCP is available to every group account with a Business subscription. Any active user in a group account can connect. There is no separate signup, approval, or access request.
The connection works both ways. As well as reading measurements, it can make changes in your account when you ask it to, and what you can change depends on your role in Stockpile Reports, the same as in the Web Application. Operations can keep the yard and the account in step: a request as plain as "Create a site called North Yard at 1200 Quarry Road, then add a pile of #57 stone to it" sets up the pile before your crew captures it. Finance can keep the record clean for reporting, for example by correcting a pile's name or checking the conversion factor used to turn volume into tons before the books close.
Deleting or deactivating a site can also affect what sits under it, and reactivating the site may not bring those back. A deletion cannot be undone, so be specific about what you ask for, and read what your assistant says it is about to do.
For the questions worth asking once you are connected, see what to ask your AI assistant at month end.
Step three: stand on the record
Stockpile Reports is the system of record for stockpile data. In our experience, most Business customers who use Stockpile Reports for inventory counts reconcile their books to our measurements. The volume is measured, not asserted, and the factor that turns it into tons is versioned. Your reported tonnage can be taken apart and checked.
That is not a promise about what AI might someday do with your yard. It is how the number is already used, and it is the same record your assistant reads from.
Where it goes next is yours
Where this goes next is not ours to define. As your team experiments with AI for dispatch, production scheduling or tighter yard-to-truck decisions, all of it depends on pulling numbers together from several systems. Our job stays narrow on purpose. We make sure the physical number in that mix is one you can trust, wherever your team decides to use it. You bring the other systems. We bring a number that does not move depending on who is asking.
Most of the industry will spend the next year working out whether its data is ready for AI. For your stockpiles, most of that work is already done. What is left is how often you measure, and whether your team can reach the number when it asks.