How good is agentic AI for supply chains?

REASONTOCHAIN brings together research, technology, experimentation, and practical analysis to understand what agentic AI can actually deliver in enterprise supply chains.

REASONTOCHAIN

How good is agentic AI for supply chains?

In 2025, I began exploring this question through my MBA thesis at the Technical University of Munich*. The research combined a survey of peer-reviewed agentic AI research with controlled experiments using a purpose-built supply chain digital twin.

Rather than asking what AI agents should be capable of, the research tested what they could actually do: retrieve information, reason, plan, make decisions, use tools, and remain coherent across interconnected systems.

The findings were then contrasted with the capabilities and claims being presented by technology vendors and industry analysts.

REASONTOCHAIN continues that research as an independent project. It updates the findings as models, frameworks, and evidence change. The standard is the same: working architectures, and measured outcomes.

I have spent more than 18 years in supply chain operations, consulting, and technology. That is the other half of the method. Agents are judged the way a planner, buyer, or network designer would judge them: cost, service, working capital, resilience, exception handling, and the human work they create or remove.

Arturo P. Martinez

Supply Chain Strategy & Technology | Agentic AI for Supply Chains

Independent Consultant, Researcher & Advisor

Engineer, MBA

What This Blog Covers
Get in Touch
Read the MBA Thesis

What REASONTOCHAIN brings

Black and white digital illustration of a human-like figure with a computer monitor head, seated at a desk with a keyboard, in a minimalist style with dotted background and dashed lines.

One place for agentic AI

A curated knowledge base bringing together research papers, technology developments, vendor capabilities, industry reports, analyst perspectives, and relevant media, organized specifically around supply chain applications.

Digital illustration of a multimodal model diagram with various shapes and lines, including circles, squares, and dashed lines, labeled 'MULTIMODAL MODEL [V2]'.

Research you can actually use

Original research, experiments, benchmarks, and case studies, including my past and ongoing work, presented with results, limitations, and practical conclusions for supply chain professionals.

Black and white multicolored digital illustration of a robot with a robotic arm on a gray background with dashed line patterns and the text 'MULTIModal MODEL [V3]' at the top.

Independent interpretation

Research is only useful when it can be translated into decisions. Interpretation of academic and technical research through a supply chain lens, what it demonstrates, where it falls short, and what it means in practice.

*REASONTOCHAIN is an independent research and knowledge project by Arturo P. Martinez and is not affiliated with, sponsored by, or endorsed by the Technical University of Munich (TUM).

Follow me on social