Question: At the beginning of the year, our management held several meetings to discuss measures to reduce costs in the long term in view of rising expenditure. In addition to energy costs, personnel costs in particular are a growing problem. We cannot do without them completely, but we are striving to make savings with the most cost-intensive employees - the specialist staff. The idea is to replace specialist knowledge with the use of artificial intelligence, databases and specialist companies in order to largely dispense with highly qualified personnel. In future, only one master craftsman or technician is to be employed, instead of the current practice with several master craftsmen, technicians and surface coaters. This would have the additional advantage that we could largely dispense with training and further education measures as well as the time-consuming search for specialists. In an era of ChatGPT and other AI technologies, the implementation of such an endeavor should be less problematic today. How would you suggest approaching such a project?
Answer: Cost reduction has always been a key issue in the industry, and consequently in electroplating. There have been periods when new ideas and technologies have not only led to more efficient processes, but also to mass redundancies. In electroplating, examples of this are the elimination of numerous mechanical processes (grinding, polishing, etc.) through improved processes such as leveling and brightening, electropolishing and the like, or generally the introduction of electroplating machines and later robot technology.
Cost reduction has always been a key issue in industry, and consequently also in electroplating technology
It makes sense to take the next step with the help of current AI technologies. However, this should be thoroughly thought through from start to finish to avoid irreversible damage from an ill-considered attempt. In the following, we highlight a few perspectives that seem particularly important to us. They are very general. If you look at your company, you will certainly come up with more specific points.
The current state of AI
While a lot has happened since our report, this technology is still in its infancy [1]. One major change since that article is the numerous new AIs that have sprung up like mushrooms and some of which are open source. This has the advantage that anyone can experiment with them and, if necessary, create their own AI that meets individual requirements.
However, this is largely theoretical. In practice, such programs currently have the status of an assistant that occasionally suffers from hallucinations. Attempts are being made to combat this with new approaches, but as far as we know, there is still no truly error-free version of the general chatbots that can always be relied on 100%.
Another problem is that electroplating in particular is an extremely specialized field. Even the technical literature is very limited compared to other branches of industry. Even information from the Internet is patchy at best, and often fundamentally wrong. This leads to correspondingly poor training data, which is needed for the development of AI based on neural networks, and therefore to poor results.
In order to achieve a usable result, you need an AI model that you first have to train with your own data. This requires you to have such a data set yourself. The practical implementation is then that you either rely on an open source solution and procure corresponding high-performance computers, or rely on an external solution such as OpenAI and others. However, the creation of a suitable data set alone will take a lot of time - and therefore money - as the selection and processing of the required data must be carried out by experts. This is hardly feasible for a smaller company. They would therefore have to rely on other partners.
Dependencies
If everything actually works as desired at some point, this can be a great benefit for a company. However, the dependencies should also be considered. Although you are currently dependent on your specialists to a certain extent, this would intensify and shift.
Not so long ago, the trend was to outsource expertise to suppliers
Not so long ago, the trend was to outsource expertise to suppliers. Plating shops with a hundred or more employees often did not even have a handful of specialists and the suppliers were expected to take care of as many problems as possible. This led to several problems. The technological sovereignty was no longer in-house, but external. In the event of complaints and internal problems, the electroplating plants were barely able to respond, and in some cases not at all. This did not have a positive effect on costs either, and changing suppliers also became a problem due to the dependency and sometimes long-term contracts. In addition, in your example, you are only dependent on one specialist, but all the more intensively. If you put all your eggs in one basket and one day - for whatever reason - your trusted master craftsman can no longer come to work, you have a very serious problem. It's like building a house with only one supporting pillar.
There are further dependencies due to external companies such as OpenAI and the like, and possibly external IT specialists. In the worst case, you work with people you don't know and rely on technologies you don't understand in the hope that everything will work out. From our point of view, you will have built your house on sand, given the current state of the art.
Experience and human sensors
In our opinion, what makes an expert is not just the data, facts and processes he has in his head, but his many years of experience. Without wishing to mystify electroplating technology, it still consists of more than the sum of its known parts. Some companies are currently in the process of compensating for this using AI [2]. This involves creating digital twins of production and thus training the AI. This enables it to deliver increasingly reliable forecasts.
This is associated with immense effort and - at least at present - does not serve as a replacement, but rather as an assistant. Furthermore, this technical approach is very individual and every electroplating company has to make the same effort to train such an AI with its own data. Even if all current technical possibilities are exhausted, this is not possible across the entire production chain - at least not at a manageable cost. An experienced electroplater sees many things at first glance. He touches the goods and quickly recognizes differences in roughness, appearance and weight. Replacing this in the incoming inspection alone leads to high investment costs.
The following aspect should also not be neglected: AI will learn more and more, and at a certain point it will also significantly exceed the mental capacities of the specialist staff, but what will be completely lost in the long term in this constellation is the input from outside. It is these new perspectives and experiences that make new employees so valuable.
Liability issues
What has not yet been clarified is the problem of liability, which is meant here in legal and not superficial terms. Who is legally liable if the AI makes process-related decisions that lead to a costly recall, for example? What about insurance in this regard?
In many areas, the use of artificial intelligence is uncharted territory and the discussions are stuck in legal as well as moral terms. One example of this is self-driving cars.
In the long term, the result is that human specialists are reduced to approving AI decisions
Of course, you can assume that your foreman/technician makes the final decision and is therefore liable for it as a person - as far as possible - but this puts them in an impossible situation. From a certain level of perfection, he will no longer be able to assess whether the AI's decision is right or wrong. We already have situations like this with AIs for games such as chess or Go, where even absolute professionals often don't know why the AI decision is the best and have to blindly rely on it. In the long term, this leads to the human specialist being reduced to approving AI decisions without understanding them. And in the end, it can lead to a lack of any chain of reasoning that led to wrong decisions in the event of a claim.
Technology not without people
The possibilities that current technology offers and promises are both exciting and frightening at the same time. From an economic point of view, the idea of saving on personnel costs may even be tempting. On the other hand, of course, there is also the social and societal responsibility that entrepreneurs have. For this reason, but also for the reasons mentioned above, we believe that these technologies should be introduced for people and not against them.
" Artificial intelligence should be introduced for people, not against them! "
As assistants, they can support your employees in their daily work. They are perfectly suited for stupid data acquisition, for example via sensors, and their evaluation. They can also be used to exchange information and sometimes even as a source of inspiration to come up with new ideas. They are also becoming increasingly important as a supporting tool for training and further education.
In our opinion, it always makes more sense in the long term to use technology to achieve more with the same number of employees, whereby the "more" can refer to higher productivity, but also better quality. Achieving the same with fewer costs can also be a goal, but very quickly comes up against natural limits and - in our opinion - does not offer such great prospects.
Literature
[1] ChatGPT and the limits of illusion; Galvanotechnik 114 (2023), No. 4
[2] On the way to digital electroplating; Galvanotechnik 113 (2022). No. 7
