Supply Chain AI Ain’t Easy

AI has taken the world by storm. Everyday life may not be quite as fanciful as The Jetsons envisioned, but cars are starting to drive themselves, and thanks to Siri and Alexa, everyone has access to their own personal assistant 24 hours a day. Beyond our personal lives, the world of commerce has also changed dramatically. Finance departments use AI to identify and prevent fraud in real-time, advise customers using chatbots, and greatly improve their ability to predict and assess loan risks. Marketing groups have automated the delivery of highly personalized ads, use AI to help build and polish their content (thanks, ChatGPT!), and can instantaneously tailor campaigns to meet the needs of local markets. These are only a few examples of how AI has transformed these two functions from top to bottom.

So why haven’t supply chain functions experienced the same? While some AI-first supply chain organizations (e.g., Amazon, Doordash) have proven that AI can create tremendous value, the vast majority of global supply chains are still very early in their respective journeys. Why does it seem so much harder to get AI to stick in this field? While there could be many contributing factors, here are five reasons it’s difficult to integrate AI into your supply chain:

• Supply Chains Operate in the Real World – If you have traveled to different parts of the world, you can appreciate that the infrastructure differs from place to place. The roads are not the same, the transportation speed varies significantly, the labor force expectation and behaviors differ from culture to culture, and the regulations are definitely location-specific.The need to marry the physical and digital worlds (where commerce gets initiated) looms particularly large in supply chain. While it’s possible to run financial books or digital marketing plans with global best practices, you simply cannot do the same with your supply chain. Practitioners need to tailor best practices to the local physical context, which can be an onerous process, and one that sometimes puts the global and local in conflict.

• Supply Chain Data and Processes are Fragmented – Supply chain processes have taken the better part of the last decade to come online. Even now, critical supply chain data is often spread across fragmented systems (for example, asset visibility may be available in one of several possible landing spots, whether it’s Project 44, or FourKites, or some other smaller vendor). As a result, building AI systems on top of your data is very difficult. Think of it this way, if you are a law firm specializing in mesothelioma lawsuits in the state of Kansas, you can know exactly how much you will pay for a lead on Google/Facebook/TikTok, whether that ad is seen in Kansas, New York, or Texas. In the world of supply chain, even the largest manufacturers and retailers don’t really know what they will end up paying for a truck load they want to move from Chicago to Los Angeles.

  
  

• The Supply Chain Industry is Heterogeneous –  One of the fun parts of working in the supply chain domain is that no two supply chains are alike. The context changes based on geographic reach, industry, business strategy, and more. A glass manufacturer’s entire P&L will be a function of how well its plants perform the change-over of colors on their lines, while a modern e-commerce company will live or die on the success of its last-mile customer deliveries.Of course, this is also tied to supply chain Data. Heterogeneity among supply chains also translates to heterogeneity in the data and data models that run them. It is not uncommon for the same data field in one supply chain context to be interpreted five different ways in five different companies (even within the same industry).While this keeps life interesting for supply chain professionals, it is a nightmare for supply chain software vendors. Software companies look for homogeneity to build standard software useful to the masses. It is no surprise that, until now, no single application for supply chain AI (like SalesForce for CRM, NetSuite for accounting, or Coupa for spend management) has made inroads in a large footprint of enterprises. Lyric plug #1 – We believe we have cracked the code on this.

• AI is Different – It is also important to understand that AI is unlike other existing technologies. In the B2B world, most software either translates a manual workflow into a digital one or creates some efficiencies within a preexisting cumbersome process. Take expense reporting:  the first wave of invoicing software allowed employees to submit their expenses online (manual -> digital). The next wave allowed them to take pictures of receipts, as well as automatically categorize the expense (adding efficiencies). This kind of software requires a shift in behavior, but it is not a complete one-eighty in terms of the way work is done.

• Contrast this with the implementation of AI-first demand forecasting and inventory planning. In this case, users must let go of their existing demand planning and S&OP processes altogether, and adopt new models that make automated decisions. It is a hard sell, and often a painful change. In the end, the gains will typically far outweigh the temporary discomfort. But all it takes is for one SKU to not have enough inventory when needed and the business will be ready to revert to its old ways, instead of finding ways to improve model performance for the long run.

• AI is Not Cheap –  The talent required to support an enterprise fueled by AI is difficult to assemble and requires best-in-class leadership. Companies like Nike, Anheuser-Busch and Target have spent years, $100+ million, and have undergone multi-geo transformations to create their internal AI capabilities.In the pre-Covid world, supply chains were not considered strategic, so the bulk of budgets were allocated to bringing AI to sales, marketing and/or finance tech stacks. Thankfully, supply chain functions that begin their AI transformations today don’t have to spend at the same scale (Lyric Plug #2 – new tech like Lyric will make sure of that.)  Not to mention that post-Covid, supply chain efficiency is in vogue, and companies truly understand the need to prioritize initiatives that make supply chains smarter.

That rounds out our top five. Because of all this, supply chain AI still requires a significant amount of energy, conviction, and sponsorship to deliver automation and intelligence throughout your organization. (That, and a great platform (Lyric Plug #3) that harnesses the best technology available with deep supply chain context baked in, of course.)  

We would be shortchanging you if we closed out without addressing one more reason – consider this an honorable mention.

HM. Vendor Hype – Look, we will be the first to admit that the vendor community doesn’t make it easy to understand what’s real and what’s hype in an ever-growing software ecosystem, especially one full of marketing content that mentions ‘AI-powered’ capabilities everywhere you look.  We often compare these marketing teams to those responsible for selling $195 bottles of Norwegian polar iceberg water. An AI-first company requires more than just task automation, demand forecasting enhancements, or AI buried so deep within preexisting processes that you barely know it’s there.

As an operator, you have to learn to not get trapped by the promises of the marketing messages. Your success depends on looking deeper, ensuring the capabilities you’ve been promised are real, and that whatever you buy comes with the power and practicality your organization needs to truly become AI-first. 

Here is our promise – a smarter supply chain will require less human touch and your customers will absolutely love it. Isn’t that worth working hard and thinking differently for?