Artificial intelligence (AI) can be defined as "the ability of a system to correctly interpret external data, to learn from that data, and to apply what has been learned through flexible adaptation, in order to solve specific problems".i Thus, AI offers the industrial sector the opportunity to automate a large number of typically repetitive tasks that until today were reserved for people.
With software packages such as Scikit-learn, TensorFlow or various Cloud solutions, the basic technology for this is available to virtually everyone today. From a technical point of view, therefore, the deciding factor is the availability of sufficient high-quality data.
Although it is extremely exciting to look at the topic from this vantage point, the technical view of the subject is only one side of the coin. Unfortunately, a technically successful AI project is still no indicator of its economic success. An AI system alone does not create any economic benefits - on the contrary, it generates development and operation costs.
The key to the success of AI systems in the industrial environment is to be found in classic lean management. While a distinction is traditionally made between value-added, necessary support activities and waste, today this definition must be broadened. Today, the use of AI allows us to automate support activities for which this was considered unthinkable a few years ago. It is therefore now possible to speak of automatable and non-automatable supporting activities. From an economic point of view, therefore, the basis for using AI systems should be a careful analysis of the value streams and processes.
An example of this at HARTING is the processing of service requests. This is a process that has been steadily optimised over the years. A previously required supporting activity was the assignment of incoming inquiries to the processing team. This task could be automated to a large extent by using techniques from the field of computer linguistics and a suitable classifier – this noticeably increased the value-creating portion within the process.
In the future, the challenge lies precisely in identifying comparable processes with automatable supporting activities and collecting and evaluating data to train AI systems for Automation.