Throughout the latter stages of my career I worked in Logistics. This involved spending a lot of time in warehouses and, as the senior IT Manager, I listened to a lot of talk about warehouse management systems (WMS). There is nothing inherent in running a warehouse that requires computers, as the need is to able to remember where you put the thing that someone asked you to store in the warehouse so you can get it for them when they ask you to take it out of the warehouse. A good memory or a pencil and paper will work for this. However, when volumes increase a computer can be very helpful. The first WMS effectively mimicked the pencil and paper, storing a list of warehouse contents and making this available for updates to multiple people at the same time. The latest generation of WMS, though, are far more advanced making many decisions of behalf of users. Configuring the WMS with the layout of the warehouse, the characteristics (dimensions, weights and so on) of the items to be stored, the likely volumes of items in and out, then allows the WMS to decide where to store items, how to group orders, how those picking the orders should travel around the warehouse and so on. Use of such a system can greatly improve efficiency and allow you to satisfy far greater numbers of orders using smaller physical space and less people than would otherwise be possible. So far, so good. This is a familiar story, and we see similar technology making inroads into all sorts of jobs in all areas of modern life. Tech giants, like Jeff Bezos and Elon Musk talk about the transformative power of artificial intelligence, and the press run stories such as this one about diagnosis of eye disease illustrating the march of technology.
However, I also heard a very common complaint as I visited the warehouses. The latest WMS are very demanding of data, and this data needs to be accurate. They also supply huge amounts of metrics and information about how the warehouse is performing. Every warehouse I visited complained of a shortage of staff who could work effectively with the WMS. People who were technically literate enough to understand what data the WMS needed, and to interpret the information the WMS produced. The skills shortage was drastic and real, and exists across the whole industry. In fact I came across several instances where the implementation of the systems had been “dumbed down” to make it less demanding in terms of data requirements, and easier for the warehouse managers to deal with.
Problems would arise for a variety of reasons. Anyone who has worked with computers knows the importance of being specific and accurate in what you ask the computer to do. Given the proven difficulty of getting people to correctly identify what they do, it´s not surprising that the larger the scope of the automation the more likely errors are to arise. What computers do provide is, though, is leverage: if you introduce an error using a system, the computer can magnify the impact of that error far more. In the WMS world this would often happen when new stock items would be introduced or ordering patterns would change. The WMS could respond in an unexpected way, blocking receipt of new orders or generating large numbers of exceptions, and then big problems could arise, as the system had not been fed properly with the right data. Reverting to a more low-tech approach, where warehouse managers took the decisions about where to store product or which orders to group together, could ease such problems. Warehouse managers were often loathe to return to the more highly automated approach after such an issue, preferring the flexibility of people versus the efficiency of system.
This is not unique to warehousing. The recent story regarding the Thai junior football team who became trapped in a cave system and their subsequent rescue shows that technology is not always the answer. Elon Musk´s advanced mini-sub was quickly rejected in favour of lots of divers and doctors. When it came to the best balance between risk, flexibility and response time a team of well motivated and skilled individuals was the best solution.
I think this tells us, despite the advances in AI, that people are still going to be needed for a lot of jobs moving forward. However, the nature of those jobs will change. The skills shortages in warehouses is not unique. People who understand how computer systems work, how to correctly collect and classify information, and how to assess the risk as well as the benefit of automating a particular activity will be at a premium. This brings us back to core decision making skills: openness to new ideas, fact collection and checking, understanding the likely accuracy of predictions and dealing with different outcomes. The most flexible learning machine of all is still a human being.