Using AI agents to address supply chain challenges and innovate sustainably

As political support for sustainability weakens, the need for long-term sustainable practices has never been more critical.

Supply Chain Analytics has traditionally been used to solve specific operational challenges (see below Four Types of Supply Chain Analytics). These methods provide the foundation for data-driven decision-making — but their potential expands significantly when paired with agentic AI.

Four Types of Supply Chain Analytics - (Image by Samir Saci)

Just like LogiGreen, which applies advanced analytics to help companies align operations with sustainability targets, organisations can now leverage AI agents to automate insights, enhance responsiveness, and scale their green initiatives more effectively across supply chains.

The following sections explore how agentic AI can transform each type of supply chain analytics, and what this means for businesses aiming to accelerate their sustainability journey.

Obstacles for green transformations of companies

At the recent ChangeNOW conference in Paris — a global gathering of innovators, entrepreneurs, and decision-makers — the mood was both determined and pragmatic. Despite a challenging context, participants shared a commitment to building practical solutions for a more sustainable future.

ChangeNOW in Grand Palais of Paris - (Image by Samir Saci)

Conversations across sectors revealed three recurring obstacles companies face when driving sustainable transformation:

• A lack of visibility on operational processes,

• The complexity of sustainability reporting requirements

• The challenge of designing and implementing initiatives across the value chain

Examples of Challenges Faced by Companies - (Image by Samir Saci)

As organisations adapt by integrating digital tools into their sustainability strategies, agentic AI is emerging as a powerful enabler. The next sections explore how this technology can help address two of the most pressing challenges:

• Improving reporting to respect the regulations

• Accelerating the design and execution of sustainable initiatives

Solving reporting challenges with AI agents

The first step in any sustainable roadmap is to build the reporting foundation.

Companies must measure and publish their current environmental footprint before taking action.

Environmental Social, and Governance Reporting - (Image by Samir Saci)

For example, ESG reporting communicates a company's environmental performance (E), social responsibility (S), and governance structures' strength (G).

Let's start by tackling the problem of data preparation.

Issue 1: Data collection and processing

Many companies face significant challenges right from the start, beginning with data collection.

Type of Information to Collect for Life Cycle Assessment - (Image by Samir Saci)

Life Cycle Assessment (LCA) — a method for evaluating a product's environmental impacts from raw material extraction to disposal —requires a complex data pipeline to connect to multiple systems, extract raw data, process it and store it in a data warehouse.

Example of Data Infrastructure for a Life Cycle Assessment - (Image by Samir Saci)

These pipelines are designed to generate reports and create harmonised data sources for analytics and business teams. Yet, navigating this complex data landscape remains a significant challenge—particularly for non-technical teams.

Just like at LogiGreen, where AI agents are being explored for text-to-SQL applications, organisations can experiment with natural language tools that bridge the gap between technical systems and business users.

Text-to-SQL applications for Supply Chain - (Image by Samir Saci)

The great added value is that business and operational teams no longer rely on analytics experts to build tailored solutions.

As a Supply Chain Engineer, there is a clear frustration of operations managers who must create support tickets just to extract data or calculate a new indicator.

Example of Interaction with an Agent - (Image by Samir Saci)

An AI agent interface — such as the one illustrated above — can enable a more seamless Analytics-as-a-Service experience, allowing users to express their requests in plain English.

For instance, reporting teams can use tailored prompts to extract data from multiple tables and populate their reports efficiently:

"Please generate a table showing the sum of CO₂ emissions per day for all deliveries from warehouse XXX."

Issue 2: Reporting format

Even after collecting the data, companies face another challenge: generating the report in the required formats.

In Europe, the new Corporate Sustainability Reporting Directive (CSRD) provides a framework for companies to disclose their environmental, social, and governance impacts.

Under CSRD, companies must submit structured reports in XHTML format.

Simple Example of an xHTML report that is not compliant - (Image by Samir Saci)

This document, enriched with detailed ESG taxonomies, requires a process that can be highly technical and prone to errors, especially for companies with low data maturity.

AI Agent for CSRD Report Format Audit - (Image by Samir Saci)

To support non-technical users, an AI agent has been tested to automatically audit reports and generate concise summaries.

How does it work?

Users send their report by email:

Email with the report in attachment - (Image by Samir Saci)

The endpoint automatically downloads the attached file, performs an audit of the content and format, searching for errors or missing values.

The results are then sent to an AI Agent, which generates a clear summary of the audit in English.

Example of System Prompt of the AI Agent - (Image by Samir Saci)

The agent sends a report back to the sender:

Agentic AI for supply chain analytics products

Analytics products for sustainability

Building effective analytics products (such as web applications, APIs, and automated workflows) has become essential to supporting sustainability initiatives.

A sustainability roadmap often begins with top management directing departments to establish a baseline for CO₂ emissions. For example, a supply chain team may be tasked with measuring Scope 3 emissions for the distribution chain in the baseline year of 2021.

Supply Chain Sustainability Reporting - (Image by Samir Saci)

Once the baseline is set, a reduction target is typically defined—such as a 30% cut by 2030. The supply chain department then plays a key role in designing and implementing initiatives to achieve those reductions.

Example of a roadmap with initiatives - (Image by Samir Saci)

In the example above, the company reaches a 30% reduction by year N through initiatives across manufacturing, logistics, retail operations and carbon offsetting.

To support this journey, analytics products are developed  to simulate the impact of different initiatives, helping teams to design optimal sustainability strategies.

Example of analytics products to support sustainability roadmaps - (Image by Samir Saci)

So far, the products have been in the form of web applications with a user interface and a backend connected to their data sources.

Each module provides key insights to support operational decision-making.

"Based on the outputs, we could achieve a 32% CO₂ emissions reduction by relocating our factory from Brazil to the USA."

For users without a background in data analytics, interacting with these applications can still be overwhelming.

AI agents offer a way to simplify and enhance the user experience by providing guided support and intuitive interfaces.

Agentic AI for analytics products

These solutions are now evolving with the integration of autonomous AI agents that interact directly with analytics models and tools via API endpoints.

Designed to assist non-technical users, these agents guide the entire process — beginning with a simple question such as: "How can I reduce the CO₂ emissions of my transportation network?"

The AI agent then takes charge of:

• Formulating the correct queries

• Connecting to the optimisation models

• Interpreting the results

• And providing actionable recommendations

Without needing to understand the backend, users receive direct, business-focused outputs like:

"Implement Solution XXX with an investment budget of YYY euros to achieve a CO₂ emissions reduction of ZZZ tons CO₂eq."

This approach, which combines optimisation tools, APIs and AI-driven support, helps make sustainability analytics more accessible across all teams — not just technical ones.

Conclusion

Using AI responsibly

As interest in AI grows, it is important to acknowledge the environmental impact of large language models (LLMs). To ensure sustainability, the core of these solutions continues to rely on deterministic optimisation models — providing reliable outputs without the unpredictability of generative systems

Large Language Models (LLMS) are used only when they provide real added value, primarily to simplify user interaction or automate non-critical tasks.

This allows to:

• Guarantee robustness and reliability: for the same input, users consistently receive the same output, avoiding stochastic behaviours typical of pure AI models

• Minimise energy consumption: by reducing the number of tokens used in our API calls and optimising every prompt to be as efficient as possible.

In short, a commitment to build solutions that are sustainable by their design.

AI agents are a game changer for supply chain analytics

AI agents are becoming powerful allies in helping our customers accelerate their sustainability roadmaps. With a non-technical target audience, this is a competitive advantage, as it allows to provide Analytics-as-a-Service solutions that empower operational teams.

This simplifies one of the biggest obstacles companies face when starting their green transformation. By communicating insights in plain language and guiding users through their journey, AI agents help bridge the gap between data-driven solutions and operational execution.

illuminem Voices is a democratic space presenting the thoughts and opinions of leading Sustainability & Energy writers, their opinions do not necessarily represent those of illuminem.