The agency aims to increase the accuracy of demand planning and forecasting from 60% to 85% of the current baseline.
Anastasia Obis
The Defense Logistics Agency, the organization responsible for supplying everything from spare parts to food and fuel, is turning to artificial intelligence and machine learning to solve the age-old problem of predicting what the military needs in its inventory.
Demand planning accuracy currently hovers around 60%, but DLA officials aim to leverage AI and ML tools to raise that baseline to 85%. Improved predictions ensure that services have access to the right items when they are needed.
“We’re about 60% accurate about what the service asks us to buy versus what’s actually on the shelf. So some of that is either we’re over-buying in some capacity or we’re under-buying. That doesn’t help prepare our systems,” Maj. Gen. David Sanford, DLA’s director of logistics operations, said at the AFCEA NOVA Army IT Day event on Jan. 15.
Rather than relying primarily on past purchase data, the model incorporates a wide range of data that DLA has not previously used for forecasting. This includes consumable consumption and maintenance data, operational data collected from war games and exercises, and data that affects storage locations such as weather.
Models are associated with each weapon system, and DLA continually evaluates and adjusts the models as they learn.
“We’re using AI and ML to bring in data that we’ve never seen before, and that’s now feeding into our planning models. We’re building and training individual models, and those become future predictive models,” Sanford said.
Some early results are already showing visible improvements. For example, the Army’s Bradley Infantry Fighting Vehicle’s predictive accuracy has improved by about 12% over the past four months, a senior DLA official told Federal News Network.
The agency has made the most progress working with the Army and Air Force, and is working on “some final data interoperability issues” with the Navy. Cooperation with the Marine Corps is also underway.
“The Army uses a lot of its sustainment data in Army 360. We’ve done a really great job of bringing it into a platform called , and now we’re putting live data into that platform and now we’re able to receive that live data into demand planning models as well as the Army. We’re on our way to the Navy, and the Air Force is next. We don’t have the precision we need, so this is our path to achieving that precision in each service,” Sanford said.
However, demand forecasts vary widely depending on the service. DLA officials cautioned against comparing predictive performance directly.
“When comparing services from a demand planning perspective, it is not an apples-to-apples comparison. Each service has different products, policies, and complexities that impact planning variables and outcomes. Broadly speaking, DLA is working with each service to improve preparedness and forecasting,” DLA officials said.
The agency is also using AI and machine learning to improve how it measures actual management and production lead times. By analyzing years of historical data, the tool is able to identify and factor in the actual performance of an industry rather than the expected delivery time. Convert to DLA inventory level.
“When we put in a request, we need the information to come back to us right away. And when we get the information back too quickly, they hold us accountable. And when it comes to production lead times, they’re not as accurate as they really are. We have what’s advertised, but what we’re actually getting, we’re not meeting the goals that we originally contracted for,” Sanford said.
Copyright © 2026 Federal News Network. Unauthorized reproduction is prohibited. This website is not directed to users within the European Economic Area.
