AI2 tofu email Trust-SupplyChain

AI And Supply Chain: Level Setting

Executive Summary

As we make our way through this series on AI, we wanted to take a step back and “get on the same page” about where AI and Supply Chain currently stand. We’ll briefly define AI as we understand it for the purposes of this discussion, then look at its three distinct venues of operation and impact within supply chain. We’re guessing this level-setting will help us simplify discussions about a technology that is currently being used to do everything from controlling robotics, to advising C-suite planning.

AI: What It Is, Where It’s Headed

One way to remain grounded regarding AI is to stay keenly aware that whatever magic it may turn out, it runs on only one type of fuel: data. Basically, the more data AI’s consume, and the better the data, the better the performance of the AI. So in a sense, AI is a dividend that can only be drawn when data investments have been made, and properly tended. As you probably know, AI turns data into answers via algorithms: discrete sets of instructions — ranging from simple to highly complex and interleaved — that computers can understand and use. The creation of algorithms requires a precise mixture of tech know-how and operational/business/scientific acumen to assure that the machine is “asking the right questions” to deliver answers that create value.  The most advanced algorithms have a recursive quality in order to feed another AI capability: continuous learning. In brief, continuous learning translates to a general instruction that says: “use the data you have and the data you generate, to keep getting better.” This particular aspect of AI performance that is most distinctive, and might also be the most “human-like” trait of AI’s.

AI’s In Other Industries

When we think of AI applications, we tend to think of the most recent and audacious (or successful) examples — as these tend to grab headlines. The Netflix recommendation algorithm will likely remain as a kind of AI staple, given the major role it played in Netflix’s early growth. On the other hand, the self-driving car may remain as a kind of AI holy grail, given some of the struggles to perfect this technology, and gain public acceptance. Most people also have at least some experience with customer-service AI’s — and most of us have probably found them wanting.  Still, they continue to improve, and no doubt about it, they are proliferating! NYU Business guru Scott Galloway recently published an excellent blog-entry on AI, and discussed how we’ve just passed through a “golden decade”  — with another likely to come. Galloway singles out biotech and military applications as areas of current, rapid advancement, but he also tracks-back to discuss more publicly familiar examples like fraud detection, and social media algorithms. He moves on to speculate in some detail about AI’s future, and in particular, about AI’s future as a competitor to human creativity. However — perhaps not surprisingly — Galloway stays away from subject of AI and supply chain.  And that’s where we come in…

Three Modes for AI in Supply Chain

The applications for AI in supply chain divide roughly into three buckets.  We can think of them as moving up from the most concrete to most abstract. We’ll take a closer look at each of these three levels — they’re all important and have immense potential, but we’re going to conclude with a recommendation to look more closely at one of these three — more on that shortly. For now, in ascending order of complexity, we have:
  • Execution
  • Decision support
  • Strategic
Let’s take a moment and briefly describe each level.


Executional AI refers to on-the-ground, operational applications, typically looking to accelerate workflow productivity. Executional AI’s can be used to pick-pack-and-ship from the warehouse. Executional AI would also include focused instructions from robotic motions to logistics flows. An example of the latter: keep these X and XY types of products together in the warehouse because they’re always purchased together. AI’s can also be used on the shop-floor to prevent machine over-use, and help lower maintenance costs.  They can also be used to flag when machines go down and recommend (or automatically reroute) for alternative processing.

Decision Support

This last executional example leads us up to the doorstep of decision-support type AI functions. In a nutshell, this decision support leverages AI to make better day-to-day choices. In general, these AI’s match the current situation with decisions and actions made in the past that resulted in the best results. “Situation” needs a little unpacking, because here, we’re implicitly stating that the AI has to take into account a wider variety of data inputs and context. This is where AI’s ability to deal with fuzzy problems becomes a needed strength. AI for decision support can also be used to mitigate or remedy exceptions, such as late shipments or purchase orders with exceptions. For example: You have a shipment that’s late departing. Are there other lanes you’ve used in the past that will be faster instead of waiting? The right AI, with the right data, can guide users to better alternatives and moreover, always be up-to-date on that research, and therefore immediately ready to recommend when the need arises. This same level of problem solving sophistication can be used to optimally allocate inventory. For example when inventory is low, an AI can recommend alternative DC’s to pull inventory from, based on similar stockout situations.

Strategic Guidance

AI applications at a strategic level operate with contextual cues, in a manner similar to decision support, but they’re not aimed at remediations for specific, discrete events.  Instead, strategic AI’s are tuned to offer proactive recommendations with wider spans of connection, aiming for bigger, longer-term outcomes. A good example of strategic AI input would be something like recommendations to promote a new supplier or carrier in the usage hierarchy.  Let’s say a company has worked continuously with five different partners. Based on performance over the last quarter, or the last year (or both) —  an AI can recommend using one carrier more than another. If the AI’s input stretches across different departments or companies, it might even recommend a new partner that has performed better — with the most similar requirements.

Reading Signals

The broader the context, and the more historical / use-case data that’s applied, the more comprehensive the AI can be with strategic guidance. For example, based on demand signals, such as favorable weather conditions or even higher engagement on social media, AI could recommend an increase in the demand forecast.  And within that same connected logical flow, AI could recommend a new allocation of purchase orders to more effectively spread out the supply coverage.

Where to Begin?

Our Execution, Decision Support, and Strategic categories aren’t meant to provide a comprehensive classification — but it does give us a basic framework for how to think and talk about various kinds of AI opportunities in supply chain. In terms of maturity, Execution is probably the farthest along with the most current applications. Go into any modern warehouse, and AI will be a staple of the inventory storage, staging and picking process. Strategic applications get a lot of attention, and tend to be represented in discussions about the future and its ambitious possibilities. As yet however, there are only a few examples in practice that have proven reliable. The potential is there, but yet to be realized. We’ll close by stepping out on this subject, and, as we promised, we’ll recommend a closer look at one of our three buckets: Decision Support applications. It doesn’t get as much play in the hype cycle, but our take is that Decision Support opportunities for supply chain AI are abundant. Call it a view from the trenches — but most supply chain professionals spend their days troubleshooting and firefighting. And while each individual decision they’re confronted with may not appear strategic, they certainly seem to add up. Our ballpark guess is that in flow decision making takes up about 40% of all daily activities within supply chain management. Furthermore, our sense is that technologies focused on middle-level complexity are continuing to improve, so Decision Support appears to be on the verge of major breakthroughs. Keep an eye on this area — we certainly will be! For more information on these topics: Check out the starter guide for AI in supply chain If you’re in IT, see how you can help your colleagues in supply chain management!
David Blonski

David Blonski

Related Posts

Responding To Uncertainty

Four Critical Steps That Will Separate Supply Chain Leaders as Recession Looms In Part One of this blog post, we discussed in some detail how

Read More >

See how Enterprise Service Management can help your business