Agentic AI and Supply Chains: What GovCons Should Know
Guy Beougher became an expert on AI after having an epiphany while working for the Defense Logistics Agency. DLA about five years ago was using chatbots, he said, but they weren’t yielding a reduction in labor despite automating human processes.
Beougher understands the power of AI and what it can do to improve Department of Defense supply chains. Now a senior advisor for Seekr and a vice president for DOD and federal logistics and supply chain at Cypress International, Beougher will speak about agentic AI and transforming supply chains during a panel discussion at the Potomac Officers Club’s 2025 Army Summit on June 18. Secure your ticket today to learn how DOD is leveraging agentic AI to improve how it procures parts and components and bolster its supply chain performance.
Ahead of his event appearance, Beougher sat down with the Potomac Officers Club to discuss how agentic AI applied to military supply chains could save DOD money; autonomously monitor, adapt and optimize supply chain workflows; and proactively address threats and vulnerabilities in contested supply chains through real time data analysis.
POC: Can you explain agentic AI and how it applies to the Army and supply chains?
Beougher: So I’ll start with where AI is going to most benefit the supply chains, not only in the Army, but for the Defense Logistics Agency. It’s in three groups. The first one is demand planning and forecasting. Right now, there are a lot of models that exist that only take history into consideration. There’s not a lot of models that try to predict the forecast, based on operations tempo, budgeted by weapon system. So we’ve consistently created a science of looking in the rearview mirror, but AI has to bring in what we think we’re going to do and execute from a training aspect or a deployment aspect. This is because, a lot of times, the AI that we need to use will have to bring in the different operating environments in which the Army will find itself.
The second group is optimizing work processes that are executed within the supply chain. Some of them are pretty simple. You could do it with bots, and I know DLA has come up with hundreds of bots that would automate a manual process that currently exists. It could also change current business processes. I think there’s a lot of application, and I think in that world, you have a human dimension that you’re fighting against as you try to optimize workflows.
The third one is really a fallout from the second, which is to use AI to become auditable in a much quicker fashion. A lot of times what happens is the equipment or the products that are on shelves in warehouses are inventoried once a month. That inventory number is sent to the buying activity, which has to employ resolution specialists that track by national inventory item number: How much was sold, how much was received and how much was moved from one warehouse to another. If AI could quickly automate that entire process, I think an audit might be possible, and that’s just one example. There’s many business processes that have far too many humans involved in them for a corrective action plan for a notice of finding recommendations. I think those are the three big buckets: forecasting; demand planning that optimizes not only organization, but workflow; and then finally applying AI to the audit process in DOD.
POC: How could agentic AI, as applied to military supply chains, save the government money?
Beougher: There are about 6 million national inventory item numbers and about 2 million of those are active, which means something was ordered against it in the last 10 years. Then there are probably 300,000 that are routinely ordered, but these aren’t keeping vehicles down. It’s the things that aren’t needed as often. Due to the age of the weapon systems, a lot of the defense industry has kind of walked away from producing some of these repair parts.
As an example, 18-to-24-months is way too long to have a vehicle down. So it’s really the cost of readiness. If you could be more predictive two years out, that you’re going to need something, especially for a weapon system that needs a repair part that you’re not gonna be able to go to the auto parts store to buy, [DOD could save] not only the costs of warehousing something that it doesn’t need, but it could also decrease the cost while increasing readiness.
POC: How can government contractors use agentic AI or AI agents to autonomously monitor, adapt and optimize supply chain workflows?
Beougher: Many already do, it’s considered intellectual property. I would tell you that they probably know what is going to sell before the government does, which is kind of a sad state of affairs. Many of the supply contracts are indefinite-delivery/indefinite-quantity. So there’s a ceiling for these, like 100 hundred parts for this company to be able to provide one of those parts on this IDIQ. So they win the award, and for the company to make money, they have to really be focused on those 100 items that they’ve said they would be able to deliver within so many days, and how many they really need on the shelf.
It’s all associated with the bottom line, and in most scenarios, that’s a really small margin compared to an original equipment manufacturer that’s producing a weapon system coming off a line. I have often told folks that you could keep it as your own intellectual property, but to be able to tell the government what your best estimate is on the repair parts in their IDIQ, I think the government would actually get a better handle on what future costs are gonna be. This is because in most cases, it’s lowest price technically acceptable, a.k.a. LPTA. It’s not a best-value [contract] yet.
Dive into the future of AI in DOD at the Potomac Officers Club’s 2025 Summer Series of U.S. military service-focused GovCon conferences! Attend three summits—the 2025 Army Summit on June 18, the 2025 Air and Space Summit on July 31, and the 2025 Navy Summit on August 26—for one low price. Stay on top of future DOD requirements for AI and register for these three premier GovCon events today.
POC: How can agentic AI enhance decision making and resilience across DOD supply chains with AI driven insights?
Beougher: It’s pretty simple. It’s having the right thing available in the right place on the globe. There are 24 distribution centers across the globe that are owned by DLA, with at least as many, if not more, of the military services having their own warehousing capability. Many times, they’re on the same installation. The thing that I haven’t talked about is using AI to determine where certain material needs to be stored and whether you should buy more, so you can store more forward to reduce the costs of movement during wartime.
There’s also another application besides demand planning and forecasting. You could build that in. But, many times, the economic order quantities would find DLA or the Army putting material in one place. When you look at the total cost, it should probably be in a few places to reduce movement during a high demand period, especially during a war.
POC: How can agentic AI proactively address threats and vulnerabilities in contested supply chains through real time data analysis?
Beougher: Agentic AI can do several things that would help perform due diligence associated with supply chain execution in a contested environment. There are many things that can contest a supply chain. One is the economic viability of a company that is producing the product that the DOD is trying to buy. What do I mean by that? We have an example from a few years ago where the sole manufacturing, I believe it was for dress uniforms up in the northeast, had a privately owned company that was the sole source for a couple of these uniform items for a couple of different services. A car manufacturer came to town, paid more money and the entire workforce left. AI could have determined that the risk was a couple years out and wouldn’t happen overnight. That could have been gleaned a couple years out so that [DOD] could start to look for more sources of supply and incentivize somebody to come into the clothing textile industry. That’s domestic because of the Berry Amendment, but that’s just one example.
I think another example is using AI to determine if there’s criminality or nefarious, counterfeit business that’s going on that’s a risk to a supply chain. As the number of vendors has shrunk in the United States, probably by a magnitude of 15 to 20 percent for DOD hardware business, many more vendors require a little more scrutiny than they get right now. There are numerous sole source small companies that may very well be at risk that we don’t know about.
I think another opportunity for AI in a contested environment is evaluating no-bid contracts. Right now, DLA has about 18,000 proposals over the last three years that have received no bids. I think DLA could use AI to better determine who to pinpoint to build a vendor base back up for those proposals that aren’t being met by industry.
POC: It seems problematic that DLA has all those solicitations out that aren’t receiving bids from industry.
Beougher: It’s relative. DLA awards maybe 9,000 bids a day. So a lot of its activity is an auto award. When you look at 9,000 a day and you have 18,000 over three years, it’s that one digital needle in the haystack that’s the problem and that’s where you need AI—what five are most affecting readiness today? The pure volume creates its own content station and I think many industries are a lot farther than DOD, to be honest with you.
POC: When I think about agentic AI and Army and supply chains, I’m thinking of aircraft parts, truck parts—things like that. Many members of the Potomac Officers Club are IT government contractors in northern Virginia. Is there an angle to agentic AI and military supply chains and IT service providers?
Beougher: That’s kind of a tender issue. The Army, about two years ago, made a decision to place all business systems budgeting under the view of logisticians. Just this last year, the decision was made to take all those dollars and that decision authority and move it from Army G-4 to G-8, which is basically buying end items. Some of it is an end item associated with software. Because of that, we were getting ready to create an enterprise resource planning system called Enterprise Business System-Convergence that would knit all of those business systems together, whether that be finance, the wholesale level, or the tactical level. I’m pretty sure that has come undone with this move. I still think it belongs to PEO Enterprise, but it was being created at the risk of taking sustainment dollars for what I’ll call legacy systems and investing them into this new ERP.
I think that’s been derailed a little bit because, right now, you have a separate business system that tracks all your money and it is Financial Management and Comptroller and the Army Budget Office that relies on that. But one of the biggest drivers is Global Combat Support System-Army, which is at the tactical level. So imagine hundreds of supply rooms being able to order on a tactical unit, obligating dollars that are being tracked by the financial automated system. Then you have a host cell system called Logistics Modernization Program that Army Material Command uses that isn’t seamlessly integrated with GCSS-Army.
When you start to talk about data and how AI can have an advantage, especially taking the human decision-maker out of it, you need pools of data that are reconciled with each other. It’s really hard to get all of the data officers, and I’m not just talking about chief information officers, there are guardians of data in each one of these institutions in DOD, and it’s really hard to get all of the data available to one source of agentic AI. That’s a challenge.
Category: Articles
