IBM: ‘Generative AI spells the end for control towers’


Thanks to generative AI, professionals can obtain answers to supply chain issues without needing to consult complex databases and dashboards. To gain more insight into the opportunities and threats associated with this technology, the Benelux department of the Council of Supply Chain Management Professionals organized an informative session in the Dutch city of Breda in collaboration with Logistics Community Brabant. During the session, IBM told attendees that it has been using generative AI for its own supply chain for several years: “We no longer need the IT department to translate questions into SAP queries.”

By Marcel te Lindert

In this summer’s Gartner Hype Cycle, generative artificial intelligence (AI) was positioned on the ‘peak of inflated expectations’. The usage of this technology in the supply chain is subject to similarly high expectations, according to a recent poll of CEOs by IBM. When asked about the functions for which they will use or develop practical applications of generative AI in the next 12 months, customer service (93%) and software development (90%) ranked first and second, but were closely followed by manufacturing (89%) and supply chain (87%) in third and fourth place. Over the coming year, it should become apparent whether those expectations can be met, or whether generative AI – like many other hypes – will end up in the trough of disillusionment.

There still appeared to be some scepticism among the audience during the Council of Supply Chain Management Professionals (CSCMP) session. As anyone who has ever submitted a question to ChatGPT knows, the answers often look reliable but sometimes turn out not to be true. “That’s why I recommend thinking carefully about the tasks for which you want to use generative AI,” stated Thorsten Schröer (pictured), Director at IBM Technology. “ChatGPT and similar tools are fine for creative tasks, when mistakes don’t matter that much. But if it’s of operational importance that you can trust the results, it’s better to use a professional generative AI platform.”

Frequent and complex processes

IBM is using generative AI in the assembly of its IBM Z mainframe computers – an activity that accounts for US$15 billion in sales. It also has a complex supply chain, because the mainframe computers require a total of 84,000 different parts from 2,000 suppliers. “Five years ago, we started to analyse all the processes in this supply chain to ascertain the frequency and complexity of each process. Those two factors determine whether it makes sense to deploy advanced technology such as AI for the particular process,” Schröer explained.

According to IBM, processes that occur only sporadically should be initially ignored because they do not merit the deployment of expensive technology. “The benefits don’t outweigh the costs,” he said. “This leaves the frequently occurring processes. If they are simple processes, you should fully automate them so that they no take up valuable time. And if they are complex processes, that’s when you should use complex technology. Those are the processes that AI can help to enhance.”

Questions in everyday language

For such processes, IBM is deploying its own AI platform called Watsonx – the successor to supercomputer Watson that caused quite a stir 12 years ago. Using everyday language, an IBM employee can ask Watsonx which orders need their attention. This generates a detailed answer, including the ability to click through in orders to access detailed status information. Then, based on that information, Watsonx provides concrete advice. “In effect, we have overlaid our SAP system with an AI layer. We no longer need the IT department to translate questions into SAP queries,” Schröer explained.

Schröer described generative AI as the “secret sauce” for supply chain excellence. “Who currently uses control towers in their supply chain?” he asked the audience. “Before long, they won’t be needed to gain supply chain visibility. Today’s control towers are too static and difficult to adapt. Generative AI enables planners to access all the information themselves simply by asking the right questions. I firmly believe that generative AI spells the end for existing control towers.”

Aggregating forecasts

IBMIBM is not alone; Amazon Web Services (AWS) is also working on generative AI. In his presentation, Felipe Chies (pictured), Senior Business Development specialist at AWS, gave examples of other applications, such as aggregating forecasts. “Or think about measuring supplier performance. Thanks to using generative AI for this, we no longer need to develop complex dashboards,” he said. “Another use relates to overall equipment effectiveness. Generative AI can explain what the problem is without the user having to analyse data themselves. The message might say that a particular production line keeps failing because of a problem with a particular machine.”

Chies emphasized that the use of generative AI is not entirely risk-free. For example, it can provide incorrect information in situations in which the system can’t possibly know the answer, or it might generate answers using data without the owner’s permission. Moreover, outcomes can be negatively affected if the data is contaminated with biases, for example. Schröer acknowledged those risks. “Research shows that 80% of CEOs have some concerns. For example, 46% are concerned about security and ethics, while 42% don’t believe that generative AI can be trusted.”

The need for smaller models

At the same time, Schröer is concerned about the sizable environmental impact of this technology. “ChatGPT requires one litre of water for every 25 questions we ask, and its electricity consumption is huge too. We can’t keep using generative AI in the current way; we will have to use smaller models – ones that may be 2% less accurate, but also consume 80 to 90% less energy. Another solution may lie in quantum computers. Their computational power is increasing rapidly: from 4,500 qubits in 2025 to 100,000 qubits in 2033,” he predicted.