DoD building AI tools to predict, mitigate risks to sustainment supply chain

In the U.S. military, as in the private sector, the logistics business is foundationally a data business. Consequently, Defense officials think it’s one of the next logical areas to begin leveraging AI to help commanders inform their decision-making.

In January, the Defense Innovation Unit awarded two contracts to start building prototypes for DoD’s forthcoming Joint Sustainment Decision Tool. One of the main goals is to use AI to move the military more in the direction of predictive logistics, anticipating potential challenges and proactively planning around them before they happen.

DoD plans to test the tool with — and is already getting feedback from — logisticians at U.S. Pacific Command, the Defense Logistics Agency, U.S. Army Europe and Africa, and U.S. Northern Command.

Lt. Col. James Toomey, the future operations branch chief in the logistics directorate at Northern Command, said the hope is that AI can help with a longstanding challenge in military logistics: the need to synchronize vast swaths of data spread across numerous DoD systems and use it for effective logistics planning.

“Really what JDST has been helping us to do is speed up that process — and without human bias, bring out the ‘so what’ of the data, and then not only look at where we have shortfalls and gaps or vulnerabilities, but then also provide recommendations and solutions,” Toomey said in an interview with Federal News Network. “So it’s really helping us get from reactive to predictive when it comes to logistics planning.”

Making logistics data more manageable

The tool is meant to serve as a central point that can aggregate logistics-relevant data from the military services and functional combatant commands, like U.S. Transportation Command, and then apply AI to analyze those raw inputs for a more complete picture than human analysts can deliver — and to produce that picture more quickly.

Part of the complexity comes from the fact that every “class of supply” — from food to clothing to fuel — has its own complex supply chains. And there are also points in those supply chains that overlap, so that a single disruption can affect multiple classes at the same time. So, for instance, finding those points of convergence in advance is a major goal for logistics planners since they’re major areas of risk.

“When you look at a deployment plan, you’re looking at thousands of rows of data,” Toomey said. “It’s hard for a human to necessarily infer in time and space all the moving pieces that go into that, and where there might be some conflict. The AI can do that very quickly and can flag recommendations for the planners or the senior leaders to say, ‘Hey, I think you might have a problem over here,’ or ‘You have some reaction time that, if you act now, you could set conditions to mitigate risk at this node or location.'”

Reducing bias

Besides stepping up NORTHCOM’s ability to quickly deal with large volumes of data, Toomey said JDST may also help reduce unhelpful bias in the logistics planning process.

“When you have groups of people putting together courses of action, if they’ve put their blood, sweat and tears into building that course of action, we inherently want to win,” Toomey said. “What the AI does better is really call out those pros and cons objectively, as opposed to having any kind of bias associated with it. The AI will dive deeper into the COA than maybe a group of personnel that don’t have all the subject matter expertise they may need to evaluate it. They may only get surface deep on it, whereas the AI can find deeper issues, or potentially surface some pros that they weren’t thinking about with the COA.”

In addition, NORTHCOM is looking to build a “sandbox” environment via JDST that can help planners think through how hypothetical scenarios would affect operations, isolated from real-world data streams.

“The scenarios that environment needs to handle range from decisions about the timing and modality of force deployment — whether supplies move by air, ground, rail or sea — to more disruptive contingencies like natural disasters that knock out a road network, rail line or port,” Toomey said. “The AI can assess that and spit out recommendations and solutions in a matter of seconds. And if we were looking at it with a group of folks sitting around a table in an Excel spreadsheet, it would take us days or weeks to really identify every level of impact and then come up with an alternate course of action.”

Prototype feedback

As one of the large DoD logistics organizations involved in the initial prototype, Toomey said he and others in NORTHCOM’s J4 directorate have been giving feedback to DIU and its vendors “almost daily.” That feedback will continue until at least March of next year, when the initial period of performance under the JDST prototype contracts is scheduled to finish.

A key priority in that feedback has been making logistics data more accessible to non-logisticians. Toomey said that data is often opaque to operators and senior leaders who aren’t experts in the field, and the tool needs to present information in a way that is immediately digestible.

“I don’t need a 25-step click path to be able to do a simple function,” he said. “I need someone to be able to look at it, to understand what they’re looking at, and then have some very basic functionality built in where they could pick it up and click around for a few minutes and kind of figure out what it is that they need to be inputting to get an output.”

Copyright © 2026 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.