In a recent report by McKinsey & Company titled: Launching the journey to autonomous supply chain planning, it is recommended that first you build a Control Tower and put a high level executive in charge of it. This will then become the foundation of “autonomous planning.” Needless to say, that any advanced S&OP function gives the kind of visibility that a so-called control tower offers. However, a control tower or S&OP solutions are far from accurate representation of the supply chain. The very first step to any autonomous planning is to have an accurate and digital model of the supply chain. S&OP does not do that, control tower does not do that. That is the reason why they are used at best for monthly planning and visibility. For accurate representation of the supply chain, as Gartner defines, one needs S&OE as the foundation for digitalization.
Once a model is accurately established, then we use AI/ML or any other kind of intelligence to run the model autonomously. Adding decision making or any kind of intelligence to an inaccurate model is like putting a very intelligent but blind traffic cop in charge of a major road crossing! He cannot visualize the world (model), he has no comprehension of what is going on and yet he is intelligent! How can he make the right decision in the absence of knowing the number of cars, their speed and direction, the length of time they have waited, emergency vehicles etc? Disaster!
The article goes on further saying that by adding more data and analytics we can then send alerts to users and examine a few scenarios and recommendations. In any, even small, model of the supply chain there are millions of variables. Systems are much better in finding optimal solutions and optimizing. Furthermore, examining a few scenarios where there are billions of possibilities, is not an optimal approach. Our understanding of autonomous is for the system to sense, act and learn autonomously and touch-free. It is done simply by detecting trends and predicting future risks and their probabilities and acting on correcting them pro-actively. Systems can do that. Systems can also respond in real-time using real-time data just like a sensor in an autonomous car and reacting and learning on an on-going basis. The definition of autonomous is not that every time the car gets close to an object, we send an alert to the driver; or let’s try a few routes and present to the driver. System finds the best route based on time and cost and takes you there. For more information on autonomous planning and use of intelligent agents in supply chain refer to Here.