The modern supply chain is undergoing a profound transformation driven by rapid artificial intelligence (AI) advancements.
During the Oracle Cloud World Tour in March, Anthony Marshall, Senior Research Director, Thought Leadership at the IBM Institute for Business Value (IBV) shed light on this pivotal shift, citing early findings of the IBV’s study on agentic AI in supply chain operations.
During the presentation, he highlighted the growing adoption of AI, particularly Agentic AI, and its tangible impact on revenue growth and operational efficiency.
Explore related questions
The Distinction Between AI Assistants and AI Agents
“The evolution [of AI in supply chains] has moved from RPA into assistance, assistance into agents, and agents into agentic platforms,” Marshall said.
The study distinguishes between traditional AI assistance and the proactive nature of agentic AI. As Marshall explained, with AI assistants, “you’re making the decisions; an agent takes all that complexity away from you.”
AI assistants primarily function as reactive tools, supporting human users by responding to queries and completing initiated tasks. In contrast, AI agents work proactively and autonomously to complete complex tasks and multi-step processes 24/7. Marshall explained that these agents leverage cognitive abilities like decision-making, self-improvement, feedback-loops and contextual awareness to adapt to real-time events.
The Rise of AI Agents for Supply Chain
According to the IBV study’s initial findings, organizations with higher investment in AI for supply chain operations witnessed an 11.02% revenue growth in 2024, a substantial 61% premium compared to the 6.83% growth experienced by their less-invested peers.
Today, agentic AI enhances an organization’s workflows by providing real-time, personalized responses to transactional inquiries. This offers significant improvements in areas like supply chain management, bottleneck identification, and design innovation.
Leveraging cognitive abilities like decision-making, self-improvement, and contextual awareness, these agents dynamically adapt to real-world events, moving “from reaction to pro-action” by anticipating problems and generating mitigation plans.
These agents, often powered by Large Language Models (LLMs) and specialized small language platforms, can be deployed across critical workflows, such as global trade management and supplier selection, driving efficiency and resilience.
However, the study also highlights key challenges that organizations must address. According to Marshall, geopolitical risks, global trade tensions, sustainability, and ethical sourcing remain significant concerns for supply chain leaders in 2025. Additionally, data accuracy and privacy concerns remain the biggest challenges associated with generative AI in supply chain operations.
The full report, which IBV co-authored with Oracle, is due in April, but these initial insights underscore the transformative potential of agentic AI in building more resilient and autonomous supply chains.
What This Means for ERP Insiders
Hold your AI agents accountable. Marshall emphasized the importance of holding AI agents accountable “just like your employees.” He explained that this involved setting metrics and monitoring key performance indicators to ensure business impact. It also includes putting people in the most impactful places along supply chain workflows to manage and remediate AI agents’ performance against their objectives. Marshall added that organizations should also leverage objectively successful agentic AI applications as a blueprint for further innovation in supply chain activities.
Deploy agents into your ecosystem to amplify impact. Marshall recommended three strategies that organizations must adapt to fully utilize AI agents. First, AI agent personas should be developed across supply chain workflows—especially those represented by the company’s global partners. Second, map out how the organization’s AI agents will work together to optimize existing workflows, create new workflows, and extend partner and customer personalized communications. Third, engage with ecosystem partners to mutually assess and support each other in pursuit of agentic AI capabilities beyond one’s organization.
Task agents to transform data. While data remains one of the largest hurdles to effectively deploying AI agents, these tools can also help organizations overcome this barrier. Organizations can utilize AI agents to explore, create, test hypothetical what-if scenarios, and provide sensitivity analysis on the data derived from extensive proprietary data and organizational experience. AI agents can also autonomously orchestrate actions required to prepare for the most impactful and likely scenarios. To achieve these goals, organizations can develop mechanisms to measure the value of agentic-led disruption avoidance to set a benchmark for continuous agent improvement.
