In this series, our goal is to examine what the new machine learning-enabled supply chain team should do to set the stage for now and in the future. In the previous installment of this series, Part One, we began a discussion of how AI/ML can help companies become more resilient and deal with change and uncertainty. Here in Part Two, we look at how AI/ML will be used to enable continuous planning and execution, as well as autonomous supply chains.
Continuous Planning Through Execution
A key element of 7x24-always on business models is that the boundary between planning and execution can become superfluous—an artificial construct we had in the past due to long information cycle times that forced us to just make decisions and commitments with incomplete data. That commitment led to that duality again—either a building up of just-in-case inventory and/or late shipments.
Star Trek’s Replicators, Creating Things With Zero Lead Time,
Are Not Available to Manufacturers … Yet
Unfortunately, creative just-in-time product does not have a replicator as the manufacturer. A longer product lead time—and that is defined as longer than the customer-expected fulfillment time—means carrying some inventory.
As we know, not all demand is predictable and supply lines may be stymied. Helpfully, though, we are constantly getting streams of information that present opportunities for sales, optimizing supply lines, and/or reacting to dramatic events. We also need the ability to process and respond to that latest information.
We want this kind of ongoing demand and supply planning and execution to be part of the holistic processes and systems we should have already set up. We don’t want to be scratching our heads at the end of the month trying to reconcile spreadsheets and emails against the actual inventory due to many changes. In other words, rather than exceptions and expedites, we want to have a continuous process.
In addition, if we are setting up an information system’s environment that is one of perpetual learning, then our machine learning engine is always seeking out important, changing circumstances/trends or smarter ways to optimize. We want it working on the whole available data and a continuous freshening of information and actions accordingly.
Continuous is both a system and process capability. There is a single platform on which the demand (forecast, orders, new patterns) and supply (production and procurement) exist in the continuum, and people are not isolated, but work in harmony. This obviously has organizational implications. Today, many supply chain organizations are either one team or at least a matrix where, ultimately, all the personnel feel accountable to one another to deliver the total result.
As well, Chief Supply Chain Officer/VP of Supply Chain is an executive position—not tucked under finance or other functions (see later section on roles and responsibilities).
This way of working surely came in handy for many organizations as we shifted to a global, always-on, ecommerce-driven, transparent, and sustainable economy. And as the pandemic unfolded, supply chain was, de facto, command central to respond and adjust to the crisis for the company.
Autonomy. What’s That?
Besides continuous, another term that is being marketed in supply chain is autonomous supply chains. The concept in autonomy is that the process will operate without human intervention. This concept has created a raging debate in supply chain circles about autonomous for planning and/or execution. For companies with lots of products (and that is relative), meaning more data about your products than you can handle without computers, relying on systems to do some of the computations, analysis, and communication is already a fact. The question that you will need to answer for yourself is: how much reliance?
Besides one’s value system or belief in whether this is desirous, there is the sheer amount of work required to achieve a totally digitized, complete, and accurate supply-chain platform. This totally digital supply chain would also ultimately include multi-tier capabilities, sensing changes and coordinating a response across multiple enterprises and service providers. One would need pretty accurate visibility of the trading partners’ operations, or at least the policies and processes for an automated timely, reliable response. Practically speaking, if this is desirous, then it is quite a journey! And is it feasible?
What is feasible is the automation of many tasks that take up part of the planners’ and production team’s day, tasks that computers can do pretty well. After all, we do trust them with an extremely important and essential process, our ecommerce, to check for availability, take and promise most orders, and set things up for fulfillment with little—or no—human intervention.1
We are so inundated with information that we might like our systems to just get to work without constantly alerting us as they find a better path forward in, say, transportation routes, locating inventory in the network, and when to build product based on demand for highly reliable replenishment items. Again, we already do operate these types of tasks today—sometimes human orchestrated and others semi-autonomously.
After researching this topic with many supply chain professionals, going the total distance to full autonomy does not seem like the goal, at leastfor now. However, the idea that there is more to automate—better automation in information flows, data quality, identifying critical events, and recommending other solutions to optimize inventory or fulfillment is most desirable. And again, this level of automation should be part of the holistic supply chain platform.
In Part Three, we discuss how AI/ML systems require a new development lifecycle and skillsets.