This has often been a debate in supply chain practices. Should we manage based on historical consumption?
You probably don’t have this debate when driving your car. You look ahead and regularly glance behind to see if any danger is approaching.
It’s common sense. Advocates of the importance of forecasting have used this argument: look ahead, keep your eyes on the horizon, to adjust your course.
This argument is misleading. When you look through your windshield, you aren’t analyzing forecasts; you’re monitoring what’s actually happening in front of you — what in the supply chain would be called actual demand: customer orders, replenishment needs, and consumption.
Forecasts are derived from historical data and enriched with market information. Driving based on forecasts, in fact, is like driving while looking in the rearview mirror, often fogged up by business assumptions. Or it’s like driving with a virtual reality headset over your eyes. Personally, I wouldn’t try that…
Forecasts are useful before you get behind the wheel. They let you plan to fill up the tank, set your departure time based on expected traffic, map out your route, and prepare for forecasted bad weather. Once you’re on the road, the only thing that matters is what’s happening right in front of you.
The analogy with your supply chain is clear: forecasts are useful for planning your route and its alternatives—that’s the heart of S&OP — but then, only actual demand matters — the heart of the flow driven by actual demand.
Undoubtedly in two respects:
Modern vehicles are packed with sensors to enhance our perception of the outside world. Radars, lidars, cameras, and speed sensors power increasingly sophisticated driver assistance systems. It can be annoying at times, but there’s no denying that it makes driving safer.
A popular example is adaptive cruise control — based on the traffic ahead of you, your speed is automatically and gradually adjusted.
In the supply chain world, computing power and sensors (i.e. data) are proliferating, but driver assistance systems are still sorely lacking in most companies! For instance, what share of factories actually pace their flow to the effective cadence of their constraints – the way adaptive cruise control would? The overwhelming majority still rely on obsolete, infinite-capacity MRP logic, in-house semi-manual Excel files, and the like.
The fault does not lie in a lack of available technology. Sensors and software exist (Intuiflow is the preferred choice, of course). What is still missing is a shared understanding, guiding principles that prevent technology from being put at the service of methodological nonsense…
Artificial intelligence has a role to play in powering these driving aids, provided it is based on relevant control principles, because in the end, it’s accountable humans who remain in charge.
One application area for these driver-assistance systems in supply chain is risk identification – anticipating risks and developing tactics to avoid them or to recover quickly once they have hit.
Whether it’s stock market manipulation or a big tech fantasy, we promise: this year, autonomous driving is finally here for real!
The promises made in recent years about artificial intelligence applied to supply chain sound a bit similar, don’t they?
Progress is undeniably immense. In many cases, semi-autonomous driving makes fewer mistakes than human drivers, who can be impulsive and irrational.
In the world of the supply chain, automation is a dream for many executive committees. Some of it is already a reality, based on algorithms and, in some cases, AI. For example, auto-approval of orders based on DDMRP buffers enables substantial automation and lets planners refocus on monitoring and improving the model. “Autopilot” adaptation of buffer sizing also contributes.
However, managing a supply chain means managing the company as a whole. Having an AI autonomously run the company is likely not the best idea. We’ve seen a few experiments in this direction, which have yielded mixed results.
One example is Claudius, an experiment by Anthropic to manage an internal cafeteria. It turned into a comical fiasco, even though it was a tiny business. So don’t rush to try this at the scale of your own company…
Project Vend: Can Claude run a small shop? (And why does that matter?) \ Anthropic
AI progress also makes it possible to quickly build decent applications. Some see in this the end of pre-packaged solutions – a so-called “SaaSpocalypse.”
Technology has invaded our vehicles – Linux sits at the core of car operating systems, just as it does for many computers and workstations. With the rise of artificial intelligence, you may soon be able to “vibe-code” your own dashboard and your own driving logic for your car.
Is that a good idea – in your car and in your company – or is it just a modernized reincarnation of the proliferation of disparate Excel files and a lack of a backbone of proven steering methods?…