Ways to think about the Supply Chain of the Future: Artificial Intelligence


Since about 2012 there’s been a massive influx of VC funding that has been poured into startups to revolutionise logistics and supply chains. And, it’s not just startups that are getting in the mix, it’s established global players like Amazon, Walmart,  IBM, Alibaba, DHL, UPS, Merck, WiseTech Global, XPO Logistics and many more that are driving new technologies forward.

Supply chains are going to change dramatically over the next few decades and that change is going to be accelerated by the convergence of new technology and evolving consumer behaviour which is going to require unprecedented levels of agility and flexibility in supply chains.

The Rise of the Consumer-Centric Supply Chain

Consumer behavior has changed dramatically over the last ten years due in large part to Amazon and the Smartphone.

Quite simply, eCommerce has made it easier and faster to get stuff. Speed and convenience has largely been innovated by Amazon, who have been able to drastically change consumer expectations by providing nearly infinite selection, one click shopping, dash buttons, 2 day delivery with Prime, and 2 hour delivery with Prime Now amongst other innovations.

Smartphones put a computer in everyone’s pocket or purse. Pre-smartphones, eCommerce was constrained by location and internet access - you could only shop from a desktop/laptop. Today, you can shop from anywhere and even have your goods delivered to pretty much anywhere. In addition, other services powered by the smartphone like ride-sharing, food delivery, and social media have conditioned consumers towards instant gratification.

As a result, consumers are now more demanding than ever, and retailers that can deliver economically, fast, and friction-less, will reap the benefits. In order to be fast, cheap, and friction-less a next generation agile supply chain is required that will be heavily augmented, if not almost entirely automated, by technology.

Supply Chains got Very Good...

Relational databases developed in the 1970’s paved the way for software companies to develop products that made a global just-in-time supply chain possible. Relational databases allowed virtually any question to be asked and answered (without the need for custom programming) like "Show me all customers between 18 and 20 who purchased Product X and live in this city". Business Intelligence applications took this a step further which resulted in vastly improved inventory and demand planning practices.

Today, companies have access to real-time data about customer demand. Supply Chain control towers allow companies to have end-to-end visibility of all supply chain data consolidated into a single view and updated in real-time. It’s gotten very good for today’s environment, but it’s going to be completely inadequate tomorrow thanks to rising consumer expectations.

But they could become Almost Perfect...


There’s a common misconception that artificial intelligence (AI), along with related tech like machine learning and deep learning,  is some futuristic StarWars-like technology, when in reality many of us encounter AI on a day to day basis without even realizing it through technology like Siri, chat bots, and photo tagging.

For a simple explanation of AI / machine learning,  I like this one from from Andrew Ng, the founding lead of the Google Brain team:

If a typical person can do a mental task with less than one second of thought, we can probably automate it using AI either now or in the near future.

The fact that supply chains and logistics companies already have a vast amount of data and processes that could be turned into actionable insights rather quickly, just means that there is potential for AI to take supply chains on a journey from very good to perfect.

Artificial intelligence techniques allow for vast amounts of data to be analyzed, categorized, and then turned into actionable insights with reduced or no human input. As an example, to better understand consumer demand, AI could browse social media sentiment, YouTube views, online browsing data, weather forecasts etc. to automatically predict the probability of a spike in demand for a specific product. And, instead of a human reviewing the results and making a decision on production or inventory levels, the AI could do that automatically and replan in real-time based on those predictions. In addition, the AI can be enhanced through machine learning, where algorithms constantly analyze data and based on the insights gained from the results, adjust their logic on an ongoing basis 24/7 in order to provide more accurate analysis.

Otto, a German retailer, is already using AI in this way to predict with 90% accuracy what will be sold in the next 30 days. According to an article published in The Economist, AI has reduced returns by 2 million units a year and is able to procure 200,000 ítems autonomously without human intervention. Over time the accuracy will get closer to 99% at a store level, and then 99% at a neighborhood level, and then 99% at a street level, and then 99% at a YOU the individual consumer level. And while that may be slightly creepy, it’s probably coming. And, once we can predict what YOU are going to purchase and when, what’s next? What new types of manufacturing and distribution business models could arise if AI can accurately predict at this level?

Automation: It’s an Evolution Not a Revolution

It’s a long game, a decades long evolution rather than a quick disruption. There are countless discrete manual tasks that make up today’s supply chain that will be automated one by one, slowly over time but at a massive scale. Take long-haul trucking, many self-driving startups and automotive companies are already working on this one piece of the supply chain on the basis that highway driving is far simpler for self-driving trucks to navigate than suburbs or congested city streets. In the U.S., trucking moves about 70% of the nation’s freight and since highways account for about 80-90% of the mileage of a long haul truck, there is a highly compelling economic incentive to get autonomous long-haul trucks into operation. Even if human drivers are needed to handle the more complex tasks of city driving and picking up and dropping off loads, a fully self-driving truck could shuttle trailers from the outskirts of a city to the next city without the need for a human driver. The more complex tasks that revolve around the first and last mile will likely be supplemented by automation slowly over time. We’re probably several to many decades away from a fully self-driving supply chain.

It seems we are in this fascinating moment of time where, regardless of industry, the speed of business is far outpacing the level of insight into the supply chain. We may be collecting troves of data, but the ability to convert that data into real-time actionable insights is still limited. And, due to the evolution of consumer behavior, the inherent unpredictable nature of supply chains, and the availability of data in the industry just means that AI techniques hold a lot of promise.

If you want to do a deep dive on artificial intelligence for the logistics industry, I highly recommend this report from DHL and IBM.

I cover forward-looking topics that are relevant for everyone that will be part of the future logistics industry. You can read more articles here or sign-up for a free but shockingly infrequent newsletter. When will it arrive in your inbox, who knows!