Hungry for Data: How Supply Chain AI Can Reach its Inflection Point

Artificial intelligence (AI) in supply chains is a chicken-or-the-egg thing. There are those who extol AI for its potential to create greater visibility into supply chain operations. In other words, AI first, visibility second.

Which may have been true when pervasive, real-time supply chain visibility wasn’t otherwise achievable. But transformative supply chain AI — including vastly powerful generative AI, which creates fresh insights, outcomes, processes, and efficiencies from massive datasets — requires we flip the equation on its head. Visibility first, followed by GenAI-driven innovation throughout the supply chain.

Imagine a regional retail manager, distributor, manufacturer, or procurement officer waking on a Monday, launching a familiar AI chatbot (maybe even voice activated), and asking in natural language if their supply chain is optimized for the week. And if it’s not, asking how the supply chain can be adjusted to meet their goals. GenAI enables this interaction with supply chain systems.

But the only way a GenAI-based supply chain solution can automatically deliver such answers is if it knows the status, location, condition, movement, etc. of every product, box, case, pallet, etc. in the supply chain. And the only way it knows that is if the products themselves can automatically communicate the information without human intervention. Today, they can, through a ubiquitous visibility platform called the ambient internet of things (IoT).

GenAI in the Supply Chain

Global consultancy Ernst & Young estimates 40 percent of supply chain companies are investing in GenAI. They’ve used GenAI to map complex supply networks, run “what-if” scenarios, forecast upstream and downstream supply, develop chatbots so partners can get answers more easily, and even generate new contracts based on past or existing agreements.

In such cases, companies are training AI models on their own, historical data and what they can glean from partners. Then they’re asking GenAI to find ways to boost efficiency. But as EY analysts put it, “GenAI tools are only as powerful as their input data, so they are limited by the quality and availability of data from supply chain partners.”

The Holy Grail of supply chain AI, however, is to generate new routes, processes, product designs, and supplier lists based on real-time data — and to do it as quickly as possible (which is quicker than humanly possible). Or as one executive told the Harvard Business Review, “When there is a supply-chain crisis, the key to being competitive is to be faster at finding alternative suppliers than everyone else because everyone’s looking to do the same thing.”

This requires training GenAI solutions on vastly more — and more current — data about actual supply chain operations. Enter the ambient IoT.

Ambient IoT: The Language of Supply Chains

With ambient IoT, products, packaging, and places carry digital signatures, which are the supply chain’s real-time visibility language, eventually feeding into the large language models (LLMs) that are the basis of GenAI. These signatures are carried via IoT Pixels, self-powered, stamp-sized electronic tags affixed to anything in the supply chain that needs tracing and monitoring. IoT Pixels include their own compute power, sensors, and Bluetooth communications, allowing products and packaging to describe their journey through the supply chain in data terms that LLMs can consume. Ultimately, they represent a bridge between the physical and digital worlds, making available for the first time, supply chain data that can actually show, predict, and optimize operations.

Ambient IoT Pixels communicate data via an established mesh of existing wireless devices, such as smartphones and wireless access points, or through easily deployed, off-the-shelf, standardized bridges and gateways installed in stores, warehouses, delivery trucks, and more. In fact, with the appropriate permissions and privacy protections, ambient IoT Pixels can extend the supply chain visibility all the way to the consumer, communicating data about product usage, re-usage, and recycling, proving the basis for more advanced GenAI models.

And they send data continuously. Unlike the supply chain records used to train GenAI models today, ambient IoT data describes the supply chain right now. With this visibility, all that’s left is to implement GenAI to answer for us, “What am I seeing in my supply chain, right now?”

Real-time visibility and ambient IoT data generation throughout the supply chain could even help address one of the challenges of GenAI: that the data used to train LLMs necessarily reflects unintentional data biases from their generating sources, which often include companies’ various ERP systems.

Products traced through the supply chain with ambient IoT speak objective truth because products are, in fact, located where ambient IoT says they are there, when it says they are. And because ambient IoT doesn’t require workers with RFID scanners to track shipments, human error can be minimized.

Ambient IoT data describes exactly the route and time products take in the supply chain. And the products carry in their digital product passports data about the parties and facilities involved in their handling. If applicable, ambient IoT Pixels could add to an LLM information about temperature, humidity, and carbon emissions every step of the way.

According to EY, one area in which supply chain companies are exploring the use of GenAI is regulatory and ESG reporting. The best, most cost-effective way of collecting vast data so that GenAI yields compliant information is through ambient IoT.

From Chatbot to Automation

Day-to-day, there are two ways a marriage of ambient IoT and GenAI could benefit supply chains. First, it would allow more people in the supply chain to understand evolving situations and take active steps to optimize or correct supply chain operations. You don’t have to be a data analyst or procurement specialist to ask a GenAI chatbot about the status of shipments or query alternate suppliers, though companies will continue to need data experts to ensure the LLMs and GenAI tools evolve to yield useful results. But the democratization of supply chain analysis and inquiry could enable the quick decision-making needed to be competitive.

Second, GenAI and other AI tools can help build a bridge toward greater supply chain automation. Through machine learning, specifically reinforcement learning often found in control systems, software can be trained to make decisions that achieve better results. Eventually, for example, they could be trained to detect supply chain disruptions before they happen and automatically engage alternate suppliers or shippers. Or they can initiate predictive maintenance by determining if certain warehouse or manufacturing systems or lines may fail.

They do this by learning from large datasets, including ambient IoT-generated supply chain data.

As we’ve learned in recent years, complex supply chains exist on a razor’s edge. A couple of minor factors can plunge them into chaos. Artificial intelligence will be critical to avoiding future chaos. But to get there, supply chains need to unlock data for things they can’t currently see. Ambient IoT delivers the visibility data that tomorrow’s GenAI innovations will be built on.