How to build real-time supply chain cost monitoring with agentic AI

A new era of automated inventory, equipment and logistic decision-making for cost management awaits. Adoption could quickly ramp up.

In brief

• AI agents for supply chain autonomously understand when to initiate a task, finish all the steps and recognize successful completion.
• Intelligent automation, analytics and IoT sensors are among the technologies working together alongside humans to achieve cost management.
• But companies need to strengthen their foundations in data to truly deliver on the promise of the agentic AI era.

This article was authored by Brian Waits, Principal, Supply Chain and Operations Consulting, Ernst & Young LLP.

Facing market uncertainty and rising operational costs, executives have a powerful ally in AI agents for driving cost management and streamlining processes in supply chains. These systems autonomously analyze invoices, market trends and supplier performance to secure the best contracts, while also monitoring equipment health and tracking shipments in real time. By minimizing costly downtime and improving logistics, AI agents empower organizations to adapt swiftly to disruptions and maintain profitability amid volatility.

However, even with 48% of tech executives reporting their companies are already adopting or fully deploying agentic AI¹, many organizations still grapple with integration challenges that could limit their ability to fully capitalize on these advancements. As the supply chain landscape evolves, robust data strategies and efficient processes become more critical, allowing companies to unlock the full potential of real-time cost management and monitoring to thrive during uncertainty.

The paradox of business today is that the future is easier to imagine than the concrete steps on how to get there. It could feel like the journey to “real time” takes an eternity. But it doesn’t have to: The glide path to an agentic-driven supply chain — the tactical capabilities to develop and sharpen now to capitalize on tomorrow — builds on many of the foundations that forward-thinking companies are strengthening today.

The need for data quality and infrastructure

High-quality data is the foundation for expanding to an agentic AI effort in supply chains or any business function. Companies typically have too much data — in too many places and too many types, without being complete or consistent. Just putting together monthly dashboards for management can consume dozens of hours, and how KPIs are displayed and reported can vary based on whichever stakeholder they talk to. In AI, such fragmentation undermines the accuracy of insights and the effectiveness of AI-driven decisions.

One effective approach is to start with only the data needed for a specific use case (see below), rather than “boiling the ocean” with all available data at once. This targeted strategy allows companies to concentrate on collecting and analyzing the most relevant data. By defining clear objectives and identifying the key metrics that drive those objectives, organizations can streamline their data collection processes and enhance data quality.

The overall goal should be to gain a full-picture view of the supply chain derived from all the places that it touches — an agreed-upon ground truth across the enterprise. By focusing on essential data aligned with specific AI use cases, organizations can improve efficiency, enhance decision-making and ultimately drive better outcomes.

Role of IoT as an enabling technology

Supply chains are becoming increasingly interconnected and efficient through the integration of IoT technologies — from raw material sourcing to product delivery. Sensors on equipment generate data on performance and the potential need for maintenance, while RFID tags provide accurate data on stock levels, reducing the risk of overstocking or stockouts. GPS devices offer real-time location data for shipments, which is essential for logistics performance and reducing transportation costs.

In real-time manufacturing, IoT devices can monitor machine health and performance, allowing for predictive maintenance. For example, vibration sensors can detect anomalies in machinery, alerting operators before a failure occurs. Additionally, smart cameras can oversee production lines, ensuring quality control by identifying defects instantly.

Establishing a standardized IoT platform is crucial for centralizing the capture of machine data. This platform allows for the systematic aggregation of data, ensuring that downstream AI applications can leverage it effectively. By standardizing data capture, organizations can create a unified data ecosystem that enhances the quality and accessibility of information, enabling more robust analytics and insights.

Continuous data streams provide a window into cost management, allowing organizations to identify cost overruns or inefficiencies as they occur, rather than waiting for monthly or quarterly assessments. Agents can analyze this data to identify trends and patterns, facilitating more informed decision-making. For instance, analyzing transportation costs over time can help pinpoint the most cost-effective routes and carriers. Furthermore, IoT fosters better collaboration among supply chain partners by providing a shared view of real-time data, aligning strategies and collectively reducing costs.

Preparing for agentic AI

Companies have used traditional AI for decades, including in supply chains, relying on less dynamic models that require periodic updates. But agentic AI is designed to continuously learn and adapt to new information and changing circumstances, enhancing the ability to process real-time data and respond quickly to supply chain fluctuations with very little human intervention. It draws upon your organization’s data, including from IoT and edge devices that may extend to your broader supplier ecosystem.

Strategic actions to take today to prepare for this agentic future include:

• Use AI to clean and standardize your data. AI-driven platforms exist today to refine and correct data with much greater efficiency. A unified data model consolidates and normalizes inputs across the supply chain.
• Centralize data for decision-making in data lakes for multiple file types. Data fabrics can also be of value to integrate and orchestrate data across the enterprise.
• Invest in cloud-based analytics solutions that can be easily adapted for new data sources and tools as needed.
• Develop data governance practices and establish clear protocols for AI-driven decision-making based on your specific use cases and values.
• Upskill cross-functional talent to understand both supply chain operations and AI technologies.
• Harmonize processes across the organization and business units to ensure consistency and efficiency in operations.
• Redefine the operating model to clarify where to use agentic AI and where to rely on human talent.

Special thanks to Scott Curtin, Guilherme Severino and Ayoub Abielmona for contributing to this article.