The artificial intelligence (AI) conundrum

Just how frequently can you have a conversation about business in general—or supply chain specifically—without somebody bringing up the topic of artificial intelligence (AI)?

While it’s hardly new, it has been galloping ahead at record speed and shows no signs of slowing down. Why would it? It’s probably the most transformational thing to come along since the advent of the internet.

So, the impact has been huge, right? Well, not so fast. Transformation often connotes speed of action, but changes in supply chain don’t rocket forward in the same way advances in technology do. While there are some fast-mover/risk takers in supply chain, most professionals are more thoughtful and deliberate in deciding where to place their bets.

And, of course, they’re constantly bombarded by phone calls, e-mails, text messages, and cold calls seeking mindshare to present the latest and greatest—so it’s small wonder that it takes time to sort the wheat from the chaff.

The primary challenge with AI isn’t the lack of data for it to process and learn from, but quite the opposite. The digital universe contains an overwhelming volume of information, as noted in last month’s column. Transforming this vast amount of data into actionable intelligence will require time, even with the advancements in AI.

As I mentioned last month, estimates say that only about 3% of available data is actually tapped. The enormity of this simply outstrips the capabilities of any existing systems—and people—to decipher efficiently what’s important and what’s not.

Consider that a world-class supply chain manager as being able to know where everything is at any time, be it on a ship, plane, train, truck or in a warehouse, on a store shelf or in the hands of the last-mile delivery carrier. It’s also essential knowing timely enough and accurately enough to take remedial action, should something untoward occur. True supply chain resilience counts on this and has yet to be achieved.

Long-time supply chain veteran in both consulting and industry, Sundip Naik, partner, principal, supply chain, Americas logistics leader for EY, is also a thought leader in terms of what’s going on and what’s coming.

He says there are three key considerations.

1. The historical nature of supply chain is a fragmented ecosystem—i.e., disconnected functional silos (buy, make, plan, move, customer service, etc.).
2. Metrics historically are not “cross-entity in nature”—i.e., cross-entity metrics that serve finance, supply chain, and commercial operations (these three entities are critical in running any organization). Supply chain metrics have almost always been sub-optimized for each supply chain individual function.
3. Legacy supply chains often reside on non-standard platforms/technologies, which means no common data fabric across all the supply chain data—across ERP, procurement, MFG, planning, and execution systems. Additionally, many of these systems have been implemented and upgraded at varying times over the last two decades or more.

These three points represent the largest challenge for deploying AI in supply chain that will drive value across the enterprise at speed, scale, and with high fidelity levels and outcomes.

The key to unlocking for AI to work at scale is a common transactional and master data repository.

This is a large effort as typically master data and transactional data are not all housed in an ERP, but rather ERP and all surrounding systems—manufacturing systems, demand/supply planning systems, and logistics (WMS, TMS) systems.

Today, EY sees a mixed bag of AI adoption in supply chain. Some use cases that are being deployed are as follows:

Procurement: Contract lifecycle management—be it AI reviewing and awarding contracts, setting prices.

Manufacturing: When lines break down, utilization is measured by OEE. AI will inform the operator on the root cause and recommend the actions, parts, people, tools to fix—all based on historical failures so the operator does not have to wait for a technician.

Logistics: Recommending new routing and re-routing of freight moves.

Product management: Recommendations on new product introductions and what the initial forecasts should be for inventory positioning.

Planning: Predicting demand, adjusting forecasts for variability (seasonality, external factors such as competitor pricing) and or customer trends.

The punch line is that AI is an extraordinarily powerful tool that will drive transformational change in supply chain over time; however, how much time to gain maturity and wide-spread usability remains an open question.