Modern AI technology in product configuration

New AI technologies to support the existing AI technology “Product Configuration“

There are things in a company that have to be explained to the same people over and over again. Examples of this range from “what do I have to bear in mind when selling my product to a new customer?” to “what does the value system of our company look like?”. Do you sometimes ask yourself how much work is put into the transfer of knowledge, and what opportunities for improvement there are?

E-learning systems can partially remedy the situation by providing employees with the necessary knowledge via courses. But even this does not avoid the need for repeated explanations. The reason for this is the “Curve of Forgetting”. This leads to a change in the level of knowledge over time because many things are simply forgotten in the absence of “reactivation”. It is therefore apparent that courses and e-learning systems typically lack mechanisms to retain knowledge in a simple way. Missing or lost knowledge leads to a number of problems and subsequent costs both in internal processes such as production and software development processes, and in outward-oriented marketing and sales processes. The opportunities for making use of Artificial Intelligence here are both helpful and highly promising.

Configuration systems [1] help meet the challenge of assembling complex products and services such as cars or financial services to match the given requirements. They ensure the accuracy of the quotation and prevent products from being offered that are not feasible in production. Very often, configuration systems are therefore a mechanism for checking the formal correctness of the artefacts offered to the customer.

Historisch gesehen sind Produktkonfiguratoren eine der ersten erfolgreich eingesetzten KI-Anwendungen. As early as the 1980s, the first configurators – then known as “expert systems” – were developed, and their use has continued to manifest itself to this day. In the age of “individualisation” and “mass customisation”, configurators have become indispensable.

However, in order to provide intelligent support for customers looking for products, contextual knowledge is needed, for example regarding their current situation. If the customer for components for automation systems has always purchased drives with a specific performance class in the past – and we know that they are now looking for drives for the same situation – then you are able to specifically limit the search space for suitable drives, or directly suggest the drives that were most recently ordered. With even more information, for example with live data about the drives used from an IoT scenario, it would even be possible to point out to the customer that they have purchased oversized drives in the past, and that a different performance class would be sufficient for their actual application. However, this kind of multi-layered context is often not accessible to configurators. Knowledge of this type would act as an enormously important lever for sales to identify the right solutions for customers.

Knowledge of this type would act as an enormously important lever for sales to identify the right solutions for customers.

To ensure that employees have the relevant knowledge precisely when it is needed (for example, to provide customers with the right advice), “predictive analytics” technologies could be used in the future. They foresee the situations in which the necessary knowledge will no longer be available without appropriate intervention. Such so-called “self-organizing learning technologies” [2] help to provide employees with the necessary knowledge at the right time in an anonymous way. From a technical point of view, it is “recommender systems” that monitor the users’ learning curves. On the basis of the learning curve analyses of similar users, they assess which contents and questions should be communicated to the employees at which times.

The use of “self-organising learning” promises time savings and qualitative improvements at various corporate and process levels. The positive effects of the increased “knowledge level” of individual employees are manifold, ranging from time savings for the transfer of routine knowledge by supervisors to improving the quality of implementation of “productive processes”, which then, for example, manifest themselves in the form of higher conversion rates and sales.

[1] A. Felfernig, L. Hotz, C. Bagley, and J. Tiihonen. Configuration Systems, Elsevier, 2014.
[2] www.knowledgecheckr.com