Transform Supply Chain Logistics with Agentic AI

AI has the potential to transform every supply chain process. While existing technologies such as predictive analytics, Internet of Things (IoT), and machine learning (ML) have improved supply chain efficiency and visibility, organizations still face significant challenges. Today’s supply chain practitioners must navigate complex scenarios ranging from geo-political tensions to natural disasters, while managing data scattered across multiple systems. These challenges create substantial business impacts. For example, a single missing fastener in a complex assembly delays an order by weeks, resulting in significant financial losses and lowered customer experience – even when all other processes work perfectly.

Can Agentic AI or “supply chain agents” solve these persistent challenges? In this blog, we will discuss how Amazon Web Services (AWS) Professional Services (ProServe) helps organizations implement production-ready Agentic AI solutions that transform supply chain operations.

Business Value Opportunity in Supply Chain

Generative AI is well positioned to significantly impact supply chains. According to McKinsey, total supply chain costs will reduce by 3-4% of functional costs or $290B-$550B across all industries. Due to this potential, EY notes that 40% of supply chain organizations are investing in Generative AI technology. However, McKinsey notes 90% of AI projects are stuck in the experimentation phase because they either start too small or lack sufficient resources. This indicates that while the opportunity is there, organizations are struggling to unlock the value of generative AI alone.

Generative AI has the potential to drive better business outcomes including:

• Increase workforce productivity by finding relevant insights and documents faster, thereby freeing up supply chain professionals’ time from more transactional tasks.
• Reduce excess inventory with visibility in material status and trust in underlying data, leading to fewer expedited and air freight shipments.
• Optimize decision making processes through automation and machine-generated recommendations, applying expertise, managing, or interfacing with stakeholders.

Agentic AI Systems work together to solve complex tasks

Agentic AI systems refer to digital systems that operate independently, interact, and make autonomous decisions in a dynamic environment. These systems can coordinate multiple agents and communicate with other AI systems to efficiently complete tasks, making them capable of complex problem-solving and automation. Generative AI provides the foundation for Agentic AI systems and agents – with AWS, customers utilize Amazon Bedrock AgentCore. Through logic-based reasoning and contextual understanding, agents plan actions, collaborate with other agents, and execute tasks efficiently, mimicking human logic and reasoning. As supply chain practitioners often deal with multiple systems and cross-functional teams or partners, using AI agents help organizations become more efficient and drive value.

There are a growing number of agent types to help complete different tasks, including model-based, goal-based, learning-based, autonomous, LLM, and agentic agents. These agents have different capabilities that work together in tandem to achieve a desired result. For example, let’s say a customer requested that their order be expedited. One agent looks up the status of an order, another one checks inventory, another one checks an expedite table and costs, and another agent brings the next recommended action together based on all the information. These agentic capabilities pull multiple data sources together for better internal and external customer experience. In addition to pulling together information and making recommendations, the agents can make changes to that information in a system of record if the organization chooses to allow it.

AI Agents in Logistics

Logistics is challenging due to the need for constant status updates, an ever-changing business environment, and multiple systems and data sources in different formats. Many companies solve these challenges with alerts and proactive monitoring, but these alerts lack context, do not give a potential resolution, and lack the ability to execute in one place.

As a guiding principle, it is recommended to have an AI agent (main) to have one focused persona such as a Logistics Agent, Inventory Agent, Replenishment Agent, or Sourcing Agent, with a team of agents underneath working together towards common goals. See Figure 1 for a visualization of an AI agent team working together. A focused persona is important to help the end user know which tasks the agent (main) is responsible for while helping restrict user data access and limit the amount of data the agent needs to process to achieve the tasks at hand. Specifically, within logistics, there are use cases for many different types of agents including: warehousing, quality, document generation, replenishment, customs/regulatory compliance, sourcing/contracts and internal or external customer experience. After defining a focused persona, the next step will be defining what problem(s) the agent will need to solve and how to get access to the data. In the following, we will focus on a Logistics Agent.

Figure 1: AI Agent team working together

How AWS ProServe created a Logistics Agent for A*STAR

In September 2024, AWS participated in the launch of Sectoral AI Centre of Excellence in Manufacturing (AIMfg), created by Singapore’s Ministry of Trade and Industry (MTI) and Agency of Science, Technology, and Research (A*STAR) as part of Singapore’s National AI Strategy 2.0. The first initiative under this collaboration focuses on exploring the “Future of Logistics”, for which AWS ProServe developed a Logistics Agent powered by Amazon Bedrock.

The Advanced Remanufacturing and Technology Centre (ARTC) is a research institute within A*STAR that consists of 96 consortium members including multinationals across 5 key verticals: Aerospace, Land transport, Consumer goods, Biomedical Manufacturing, and Energy. This organization drives research and development across four strategic themes:

• Next-Generation Manufacturing Processes
• Autonomous Manufacturing
• Net-Zero Manufacturing
• Resilient Value Chain

In alignment with Industry 5.0’s emphasis on human-centric, sustainable, and resilient production, A*STAR ARTC sees Agentic AI as the catalyst for empowering plant teams with virtual AI Agents with the opportunity to:

• Encapsulate institutional knowledge across planning, execution and supplier collaboration, embedding it into the organization’s operational DNA.
• Operate autonomously by making goal-driven decisions, self-improving through feedback loops, and maintaining contextual awareness.

Together with AWS ProServe, A*STAR ARTC co-developed an AI agent tailored for Logistics Specialists and Analysts. This intelligent system allows supply chain practitioners to:

• Aggregate and synthesize real-time data from Enterprise Resource Planning (ERP), Transportation Management System (TMS), Warehouse Management System (WMS) and customer-facing portals
• Deliver instant, accurate responses to internal and external inquiries—eliminating up to 50 percent of the manual lookup and reconciliation workload
• Reduce expedite costs by 3% – 5% of total logistics spend, mitigate revenue leakages, and shorten order-to-delivery cycles
• Boost planner productivity by minimizing rework, allowing focus on exception management, network optimization, and strategic supplier engagement
• Elevate customer satisfaction through rapid, transparent updates and predictive ETA insights

Beyond immediate efficiency gains, this AI-driven approach underpins a robust data strategy which positions logistics as a catalyst for smarter, more informed decision-making across the operations value chain, from capacity planning to after-sales service.

Approach to Building the Logistics Agent and Results

The AWS ProServe team worked with A*STAR to define several problems or tasks for the agent to solve such as shipping information updates and alerts for impacted purchase orders. The supply chain practitioners interact with the data using natural language and conversational AI to change, cancel, or make recommendations to solve an inquiry. Once the team had defined the different problems or tasks, we built a Logistics Agent utilizing Amazon Bedrock and other AWS services.

Video 1: AI Agent – Problem Statement to Execution

As shown in Video 1, the introduction of a Logistics Agent allows teams to get updates of information from multiple sources, such as weather, shipment updates, and more, faster, get actionable insights, and receive a standard answer to an inquiry. For example, the user requests an update on a purchase order and types in the question in natural language. The AI Agent understands the question and identifies the right data source that helps to answer the question in natural language by analyzing structured or unstructured data. This includes internal data sources such as ERP System or Excel spreadsheets, or external sources such as websites from ports or Application Programming Interface (API) connections to air freight carriers. Next, the AI Agent accesses the relevant data to answer the question with natural language processing to provide the correct response. See Figure 2 for a visualization of how the data connections and agents setup are set up.

Figure 2: Example of Agent setup with access to internal and external data

In summary, a Logistics Analyst no longer needs to manually find information and derive insights which allows them to focus on more strategic tasks. This is one example, but there are many more applicable examples across supply chain where generative AI and supply chain agents transform the way organizations operate. AI Agents help improve customer experience by deriving insights instantaneously, answering end-customer inquiries in seconds, and enable self-service inquiries.

Conclusion

Agentic AI capabilities are transforming the way logistics practitioners operate and execute their day-to-day business while improving the end-customer experience. The Logistics AI Agent allows supply chain teams to engage in natural language, understand organizational context, identify the right data sources automatically, and make conclusions or recommend the next best actions utilizing AI reasoning. Leading with business value, there are opportunities across all functions within supply chain resulting in increased productivity, increasing revenue, increasing speed, reducing cost and eliminating waste. Leaders in this space will be able to capture the value early on and turn this into a competitive advantage as end-customers are more demanding than ever.

Companies who begin their journey will realize the business value faster and quickly gain a competitive edge. AWS customers can begin building today with Amazon Bedrock suite of services and other services available. If you want help expediting your journey, please reach out your AWS Professional Services account executive or AWS Account Manager point of contact.

Special thanks to Sam Gordon for his extensive contributions initially building the Logistics Agent, Annie Naveh for further Agent development and ongoing support, and Emily O’Kelly for additional support and guidance.