Utilization of Artificial Intelligence (AI) to Illuminate Supply Chain Risk

Introduction Artificial Intelligence (AI) has been a heavily researched topic with its concept evolving over recent decades and its scope of capabilities explored at renowned conferences like NeurIPS, AAAI Conference on Artificial Intelligence, and World AI Summit, leading R&D institutions like MIT, Harvard, and CMU, and even over family dinner tables. As groups have attempted to explain what AI is and what it is not, the number of definitions and societal confusion over what it is has unsurprisingly multiplied. At its core, AI is the ability of systems and/or machines to perform tasks that normally require human intelligence, such as recognizing patterns, learning from experience, drawing conclusions, making predictions, or taking action.¹ By leveraging advanced algorithms, large datasets, and powerful computational tools, AI enables machines to learn from experience, adapt to new information, and autonomously make decisions. AI can be viewed as a broad and multidisciplinary field that encompasses several subsets, such as robotics, machine learning, deep learning, natural language processing, and more.

Robotics involves the design and creation of physical machines (robots) capable of performing autonomous or semi-autonomous tasks. Machine learning (ML) is the science of using algorithms, which are a set of rules designed to perform a specific task or solve a particular problem, to learn from data and focuses on performance improvement over time. Deep learning is a specialized area within machine learning that is inspired by the human brain and uses artificial neural networks to solve problems like image and speech recognition. Natural language processing is another subset of AI that uses machine learning to enable computers to process human language and is used for tasks such as text analysis, sentiment analysis, and language translation.

From virtual assistants like Siri to autonomous vehicles built by Tesla and advanced robotics, AI is transforming industries and everyday life, pushing the boundaries of what technology can achieve. AI assists humans in everyday life by making intelligent decisions in a short amount of time and at scale. This process is made possible by enabling leadership buy-in, adopting ethical development of AI, powered by data processing and complex programming by highly skilled data scientists, data engineers, AI architects, and ML engineers. While AI cannot reason in the same way as humans³ can, it can automate repetitive tasks, detect patterns, and provide recommendations that enable people to focus on creative problem solving.

As with all technology, though, AI has its fair share of limitations. There are operational limitations, which include a dependence on high quality, abundant data, as well as ethical limitations like bias and overreliance. Ethical limitations can be managed by implementing a Responsible AI (RAI) strategy. RAI is the practice of using AI in a way that emphasizes human oversight and societal wellbeing. It’s about ensuring that AI models, datasets, and applications are developed and deployed ethically and legally, without causing intentional harm or perpetuating biases.⁴ The Defense Logistics Agency (DLA) manages ethical limitations by utilizing an RAI approach that provides a centralized process to identify, track, and improve the alignment of AI projects toward RAI best practices and the DOD AI Ethical Principles while capitalizing on opportunities for innovation.⁵

Despite its limitations, AI remains a transformative technology for an organization with access to abundant, high quality data sources. It can be particularly beneficial for supply chain risk management (SCRM), which is a systematic approach used by DLA to proactively identify, assess, and mitigate risks throughout the supply chain. This includes addressing vulnerabilities, threats, and potential disruptions to maintain the integrity and uninterrupted flow of materials, products, and services.⁶ Specifically, AI can assist with SCRM by improving visibility via predicting bottlenecks or disruptions, forecasting customer demand to optimize supply chain workflows, and recommending alternative, pre-qualified suppliers during disruptions. Beyond these common SCRM applications, AI can assist with different types of supply chain risk. To further explore these, the authors of this paper have grouped supply chain risks into two broad categories: known and unknown.

To use common examples, a known supply chain risk could be when a supplier becomes bankrupt, which can be predicted by monitoring the supplier’s financials over time. An unknown supply chain risk could be when a squall forms and forces a container ship transporting critical, Class I level supplies to change course, thus prolonging the original estimated arrival time to warfighters. Although it is easier for AI to predict known risks, it can be used to reduce the probability of unknown risks. For example, squalls, as mentioned above, are a dangerous weather event typically prevalent in the Inter-Tropical Convergence Zone (ITCZ), which impacts DLA’s supply chains for international Area of Responsibilities (AOR) associated with higher-than-average adversary threat levels belonging to customers such as U.S. Southern Command, U.S. Africa Command, and U.S. Indo-Pacific Command. Weather events like squalls cannot necessarily be predicted but modern supply chains incorporate weather data to reduce the probability of a weather-related disruption from occurring.

AI for supply chain management (SCM) has been researched and implemented in the commercial space for years. A 2021 study analyzed AI techniques utilized across four SCM fields (marketing, logistics, production, and supply chain) and found that artificial neural networks, which are associated with deep learning, were the most common technique used from 2008-2018.⁷ One of the best and most diverse applications of AI for SCM/SCRM in the commercial space is Amazon. Amazon has been using their “Supply Chain Optimization Technology (SCOT) to forecast demand for the past decade, AI-enabled robots to recognize, sort, and inspect goods at warehouses, predictive meteorology to make safe, informed driving decisions, and route optimization for last mile delivery.”⁸ Walmart is another success story when it comes to adopting AI technology for SCM. Their Route Optimization technology enabled them to “avoid 94 million pounds of CO² by eliminating 33 million unnecessary miles driven and optimized routes to bypass 108,000 inefficient paths” in 2023.⁹

As government agencies face numerous policy barriers, they tend to trail behind the commercial space in terms of adopting advanced technology at scale. The Department of Defense has aimed to shorten this gap by publishing a Data, Analytics, and AI Adoption Strategy (DOD DAAIS) which outlined six key goals: improve foundational data management, deliver capabilities for enterprise business and joint warfighting impact, strengthen governance and remove policy barriers, invest in interoperable federated infrastructure, advance the data, analytics, and AI ecosystem, and expand digital talent management.¹⁰ The DOD DAAIS states that “accelerating the adoption of data, analytics, and AI technologies will enable enduring decision advantage, which is a competitive condition characterized by five core outcomes - battlespace awareness and understanding, adaptive force planning and application, fast, precise, and resilient kill chains, resilient sustainment support, and efficient enterprise business operations.”

By taking this proactive approach supported by a data-centric mindset and leveraging AI capabilities instead of a historically reactive and disconnected approach, DLA can simultaneously build more resilient and efficient supply chains, reduce the impact of disruptions, and increase availability of stable support for the warfighter at home and abroad.

DLA’s Journey with Supply Chain Risk Management  (SCRM) and AI

DLA plays a foundational role in supporting military operations and critical domestic needs, serving as a primary logistics provider across the United States’ Defense Industrial Base (DIB) and as an essential partner in homeland resilience. DLA’s ability to deliver reliable and rapid support to U.S. warfighters as well as Whole of Government (WoG) partners is essential. “Interruption of DLA supply chain operations compromises our nation’s ability to deliver combat power and execute critical missions.”¹¹  As part of the 2021-2026 Strategic Plan, DLA published a Supply Chain Security Strategy outlining how the agency will address supply chain security challenges across the enterprise. The strategy identifies the following four strategic focus areas:

• Institutionalize supply chain security across the DLA enterprise
• Maintain integrity and access to key data
• Partner with valid, reputable vendors who produce quality supplies and services
• Strengthen the resiliency of systems, processes, infrastructure and people.⁷

Coupling these initiatives with AI technologies will transform DLA’s logistics into an intelligent network capable of adapting to complex, rapidly changing demands and environments. DLA is fueling this AI-enabled transformation by building a data-centric workforce, investing resources to support the development of scalable AI solutions,¹² and funding Small Business Innovation Research (SBIR) projects.¹³

Further propelling DLA’s journey with AI, the agency established an AI Center of Excellence (AI CoE) in June 2024. The mission of the AI CoE is to “increase and coordinate the safe and responsible integration of AI throughout DLA, ensuring robust safeguards are established to mitigate risks”.  If applied correctly and responsibly, AI can revolutionize DLA’s operations and support DLA’s strategic vision, particularly around SCRM. Since the inception of these AI initiatives, DLA has bolstered their AI inventory. Today, DLA uses multiple Business Decision Analytics (BDA) Supplier Risk Assessment models to automate the identification of potential bad suppliers who provide counterfeit, non-conformant, or overpriced items, placing the warfighter and the agency at risk. To date, the BDA Supplier Risk models have analyzed 43,000 vendors, identifying more than 19,000 as potentially high risk. Recently, BDAgenerated information was used to trigger an investigation, which led to a suspect actor pleading guilty to providing parts made in Turkey for U.S. weapon systems. Against several U.S. laws and regulations such as the False Claims Act, Buy American Act, and Arms Export Control Act, the supplier falsely certified compliance with U.S. procurement laws while outsourcing manufacturing to Turkey. More specifically, they misrepresented that parts were domestically produced, made fraudulent claims for payment from the DOD, and illegally transferred technical data and manufacturing specifications to Turkish firms without the required export licenses.

Beyond detecting risk, the BDA model enhances supply chain resiliency and cost efficiency by filtering unreliable suppliers, ensuring that materiel meet specifications and are delivered on time. This reduces delays, prevents operational disruptions, and minimizes the likelihood of defective parts entering critical defense programs. The models help DLA avoid inflated procurement costs by flagging overpriced items, ensuring defense budgets are maximized without sacrificing quality. The success of DLA’s BDA model positions it as a template for broader SCRM initiatives across the DoD, aligning with the Government Accountability Office’s (GAO) recommendation for tighter vendor oversight.¹⁴  DLA’s efforts showcase how AI-driven analytics enhance accountability, streamline investigations, and preempt supply chain threats. DLA’s BDA models stand as a key asset in protecting the warfighter and ensuring national defense operations are fueled by reliable and compliant suppliers.

In 2019, DLA Aviation created the Long-Term Contract (LTC) Negotiations Analytics (LNA) tool as an R&D Proof-of Concept to develop an empirically driven, probabilistic modeling approach to generate contract parameter recommendations to optimize the reduction of overall LTC costs. The LTC Parameter Optimization Model ensures suppliers can effectively support the warfighter by streamlining the defense supply chain and enhancing procurement efficiency. Using past data and running advanced simulations, the model quantifies demand variability in terms of risk of over-procurement versus the benefit of quantity-sensitive unit price breaks, helping DLA make best value procurement decisions for its customers. One of the key benefits of this model is identifying materiel where the data suggests DLA can assume more risk in the form of higher order quantities, thereby driving increased supplier interest and ensuring supplies are always on-hand, so warfighters are not left under-equipped. Suppliers also get a boost from steady, reliable contracts, which makes it easier for them to invest in resources and infrastructure. It keeps them prepared for future demands and strengthens their long-term commitment to support the defense supply.

DLA’s model real-time dashboard lets suppliers adjust, helping them stay responsive to shifting battlefield conditions and new threats. Cutting waste and aligning procurement with actual demand makes better use of defense budgets, freeing up resources for advanced tech and increasing readiness. Ultimately, the LTC Optimization Model strengthens the bond between suppliers and the defense sector, ensuring warfighters have the necessary materiel and supplies when it matters the most, directly supporting their missions and keeping operations running smoothly.

While DLA is working towards advancements in the fields of SCRM and AI, we are continuing to see traditional, linear supply chain models face new challenges in an environment marked by unpredictable disruptions, contested logistics, and the increasing complexity of modern supply needs. DLA continues to tackle these challenges head on, with an eagerness to apply innovative and effective solutions. A recent example of DLA’s shift to an interconnected approach was evident through Operation Allies Welcome (OAW). OAW was an initiative directed by President Biden in August 2021 that supported Afghan evacuees, which required rapid mobilization, real-time coordination with WoG partners such as the Department of Homeland Security, and seamless responses at various CONUS locations.¹⁵ This highlighted DLA’s capacity for largescale humanitarian support as well as the limitations of a traditional supply chain in a high pressure and multifaceted scenario.

AI Capabilities and Art of the Possible

In the face of an increasingly complex global logistics landscape, DLA must ensure its supply chains remain resilient, efficient, and adaptable. The GAO recently highlighted vulnerabilities in key areas such as stockpile management and Ground Combat Systems (GCS) inventory.^{16, 17} In contrast, the Department of Defense Inspector General (DOD IG) conducted an audit and discovered that DLA did not consistently manage or provide oversight of the Defense Fuel Support Points (DFSPs) per DOD policies. These gaps threaten the operational readiness of the DOD. AI offers transformative opportunities to address these challenges, enabling the DLA to manage risks, optimize operations, and improve decision-making processes.

Tantalum, a critical material essential for military aircraft, represents a supply chain vulnerability case study. The notional supply chain for tantalum begins with raw material extraction, followed by a complex refinement process into tantalum salts, powders, and metal ingots before being incorporated into end products such as jet turbines. This supply chain involves multiple stages, each presenting opportunities for inefficiencies or disruptions. AI tools offer actionable solutions to address these vulnerabilities. For example, inventory monitoring systems powered by AI can track tantalum as it moves through each stage of the supply chain, from mining to processing and finally to its use in military products. By providing continuous visibility, AI ensures that material flow remains transparent and potential delays are minimized. Additionally, AI-driven predictive analysis uses machine learning models to anticipate demand, identify supply chain risks, and optimize transportation routes. This capability allows DLA to stay ahead of potential disruptions and maintain a steady availability of materials.

The vulnerabilities within the DLA supply chain are further compounded by significant gaps in data modeling for the DOD’s strategic and critical materials. The GAO findings reveal that the DOD has developed a process for identifying material requirements to determine shortfalls, but it does not have the appropriate data model requirements for 115, or roughly 40%, of the strategic and critical materials it wanted for fiscal year 2023. Over 90% of the materials the DOD identified in shortfall in fiscal year 2023 had zero or one domestic supplier.¹⁸  This gap hinders the DOD’s ability to assess risks, forecast demand, and plan for contingencies. Without accurate data, decisionmakers face challenges in identifying vulnerabilities or inefficiencies in sourcing and distribution. AI provides a powerful solution to overcome these data limitations. AI systems can aggregate information from multiple sources through advanced data fusion and integration, filling critical gaps and creating a unified view of the supply chain. This comprehensive perspective enables more accurate modeling of stockpile requirements and better identification of potential weaknesses. AI tools can also analyze defense-specific and civilian-use materials to pinpoint vulnerabilities in sourcing or continuity of supply. Risk detection capabilities, driven by machine learning, monitor these critical materials to ensure that disruptions are identified early and addressed swiftly.

The findings and recommendations in the document highlight significant opportunities for leveraging AI to improve the management of DFSPs. AI offers transformative solutions that enhance oversight, optimize resources, and mitigate risks across these critical facilities, ensuring compliance and operational efficiency. One of the most impactful applications of AI lies in remote monitoring and inspection. AI-powered drones and advanced sensors can perform visual and thermal inspections of storage facilities without the need for continuous in-person visits. These tools provide real-time monitoring, enabling early identification of structural weaknesses or environmental hazards. This approach significantly reduces labor-intensive manual inspections while enhancing the reliability and frequency of monitoring activities.

In addition to monitoring, AI improves predictive maintenance by anticipating maintenance needs before problems occur. By analyzing historical data and real-time equipment conditions, AI can schedule maintenance activities proactively, minimizing the risk of fuel leaks, spills, or costly breakdowns. This predictive capability ensures that maintenance efforts are focused on areas with the highest risk, extending the equipment and infrastructure life cycle while reducing unplanned downtime. AI also plays a crucial role in resource optimization, particularly in staffing analysis. By evaluating staffing data and workload trends, AI tools can identify gaps or inefficiencies in workforce allocation. This ensures that regional offices are properly staffed to meet compliance requirements, streamlining operations while maintaining readiness. Furthermore, AI enhances compliance monitoring by automating critical oversight processes. For example, AI systems can track Staff Assistance Verification (SAV) schedules and generate automated alerts for upcoming or missed inspections. These automated reminders improve accountability and ensure compliance timelines are met, reducing the risk of regulatory issues or lapses in oversight.

To support decision-making, AIdriven data analysis and reporting provide valuable insights into DFSP operations. AI tools can analyze fuel loss data to detect anomalies, such as unexpected drops in fuel levels, which may indicate leaks or theft. By identifying these issues early, DLA can take corrective action before they escalate. Additionally, AI systems generate comprehensive reports on compliance and maintenance activities, offering decision-makers clear visibility into areas for improvement and enhancing accountability across operations. Lastly, AI enables robust simulation and scenario planning, allowing DFSP managers to assess risks and prepare for potential disruptions. Various risk scenarios can be simulated using AI models, such as equipment failure, environmental hazards, or supply chain interruptions. This proactive approach provides DLA with actionable strategies to mitigate disruptions, ensuring that fuel support operations remain resilient and adaptable under any conditions.

Implementing these AI-driven capabilities will significantly improve the management of DFSPs. AI enhances oversight by providing continuous monitoring, ensures compliance through automated alerts, optimizes resource allocation, and identifies risks early through advanced analysis and predictive tools. These solutions align with the recommendations to strengthen DFSP operations, reduce vulnerabilities, and improve overall accountability and efficiency within the fuel support infrastructure.

In both competition and conflict, AI has the potential to significantly enhance the capabilities of the DLA Joint Logistics Enterprise (JLEnt) by addressing critical logistics challenges currently impacting the enterprise, including, but not limited to, munition 5200.44 (2024).¹⁹ DODI 5200.44 (2024) is an issuance associated with implementing a program for information and communications technology (ICT)SCRM. AI can support DLA with this mandate by compiling supplier level data, narrowing the focus on suppliers operating in this subcategory  (e.g., ICT), and predicting supplier risk. An end user facing output from this system could be as simple as a list of high-risk ICT suppliers broken down by parts and rationale outlining why it was marked as high-risk that is updated daily. The examples discussed in this paper merely scratch the surface of how AI will expand DLA’s capabilities and drive innovation.

Conclusion

The DLA strategic adoption of AI is reshaping the future of SCRM, addressing vulnerabilities, enhancing operational resilience, and ensuring mission readiness. AI’s integration across DLA’s logistical infrastructure, monitoring equipment conditions for predictive maintenance, and automated compliance tracking underscore its transformative potential in safeguarding national defense operations. By leveraging AI-driven tools for predictive analytics, resource optimization, and scenario planning, DLA proactively mitigates risks, ensures seamless fuel and materiel support across DFSP and other sites, and reduces downtime through timely maintenance. AI’s capacity to simulate risk scenarios, automate oversight, and identify supply chain anomalies, such as fuel loss or supplier inconsistencies, empowers decision-makers to respond swiftly and effectively to emerging threats.

DLA’s AI initiatives align with key DOD policies, including DODI 5200.44 (2024), which mandates operational ICT SCRM programs to manage supply chain risks comprehensively. As DLA continues to expand its AI capabilities through workforce development and technological innovations, it is poised not only to meet but exceed the requirements set forth by the DOD for secure and efficient logistics operations. Ultimately, DLA’s AI-enhanced logistics framework strengthens its ability to navigate complex global supply chains, ensuring uninterrupted support to warfighters, fortifying the integrity of defense infrastructure, and maintaining strategic superiority. By embedding AI at the core of its operations, DLA is positioning itself as a model of innovation and resilience, ready to face the evolving challenges of tomorrow’s contested logistics environment.