Mercedes-Benz uses generative AI to accelerate digitization

Mercedes-Benz has long used machine learning and classic AI. But now you’re also using the AI ​​you create, for example in the MO360 production environment. What exactly is it all about and to what extent does it change the job profile of employees?

– Job descriptions in manufacturing and management are changing due to the increasing use of digitalization and powerful AI systems. AI aims to improve and simplify the daily work of employees. New digital tools, for example, make it possible for production workers to improve processes and quality management in the long term. The first part results of MO360 are very promising, especially as we see it being adopted not only by IT professionals but also by foremen on the shop floor.

Your organization has launched a skills initiative called Turn2Learn. What AI or digital skills are being taught to employees?

– Turn2Learn is an initiative of our HR department focusing on digitization and AI. The offering ranges from AI and machine learning for beginners, prompt engineering, courses in programming languages ​​like Python, deep learning and neural networks, reinforcement learning, RPA and natural language processing. In total, employees have access to more than 40,000 courses on data and AI qualifications across various external educational platforms. We started a great team effort within IT because our biggest asset is our employees. Therefore, it is very important for us to attract and retain the best employees and help them fulfill their individual potential.

Does Mercedes provide training for production staff or office staff only?

– We invest in improving digital capability in all areas of the company. Whether colleagues in manufacturing or management, everyone needs relevant knowledge and new skills to effectively use AI applications in their day-to-day work. In two pilot projects, we are currently training over 600 employees from across the group to become data and AI experts.

Do you expect generative AI to lead to job losses?

– We cannot yet say with certainty what effects increased digitization and generative AI will have on future work lives. However, it is clear that work patterns will change, and so will job profiles. That’s why talent is critical to successful change.

How does the work of employees change during the digital transformation process and as part of the introduction of AI?

– Some functions will definitely be transferred to AI applications in the future. But this is a positive thing, because just as automation and manufacturing robots have changed the way cars are made, more freedom has opened up for strategic or creative work.

Speaking of training staff, how far has Mercedes-Benz come with generative AI?

– We’re actually manufacturing with buildable AI in some areas – and we’re not just talking about pilots. For example, we’ve been using Github Copilot in software development since May, and we’re seeing significant performance gains there. We also use the AI ​​we develop in the customer environment. For example, in the UK, an intelligent virtual assistant can interact with customers on a website and provide specific answers to questions about owner’s manuals and vehicle information.

– And our data platform MO360 enables a generative AI to analyze and process data in a digital ecosystem for manufacturing. And with the help of a large language model, data or data formats are available so that they can be queried by production workers using natural language, not just by experts using highly specialized database queries. We are currently testing this using Chat GPT. Ultimately, AI accelerates the democratization of data usage.

In which areas do you see the most potential for AI?

– We have worked very intensively on this issue and tested external studies and AI internally. On the one hand, we have software development. Whether it’s technology and vehicle development or the business side, we’re seeing very significant efficiency gains.

– Another part is the conversation with the customer. For the foreseeable future, AI’s direct interaction with the customer, currently being tested in the UK, will probably be the exception. But I firmly believe that AI applications will help further improve the customer experience and make processes more efficient.

– Another area to invest a lot of brain power in is parametric design in technology. There, AI can lead to huge increases in productivity as it supports people at work.

And input is possible via voice or keyboard – should employees be trained in AI?

– Initially, there is basic training for limited use cases in production. In addition, our employees have additional training opportunities in the course, including a learning pathway for immediate engineering. But they learn to use these tools creatively to test things out and see what works and what doesn’t.

– However, in general, I believe that prompting or prompt engineering is something you have to learn, so we are thinking about training in it more widely across the company, not just for selected IT and data. experts. It certainly helps to get the most out of AI.

What problems have you faced so far with AI or Chat GPT?

– Hallucinations are actually a challenge. It was a very delicate balancing act when it came to direct customer contact in the UK. You can rule out hallucinations with reliability checks and related controls, but if you set the criteria too narrowly, the machine will say “can’t comment on that” more often than you’d like. You have to be very careful and find the right balance. How to capture hallucinations is perhaps the most important question to be solved right now, and is also at the heart of AI research.

Will Mercedes-Benz train its AI tools with only its own data?

– Yes. For example, if we want to visually explain our vehicles to customers, we can only do that with our own training data. Also, training in these AI environments takes place exclusively in secure areas so that data cannot be disseminated publicly. There is also some public data that we can use for AI, but we rely on our own data, especially in a manufacturing environment.

In addition to Azure Open AI services in the production environment, what role do other AI solutions play for Mercedes-Benz?

– Open AI is currently portrayed in the media as an AI spearhead. And this is a great technical solution but we don’t limit ourselves to that. Of course, other companies have interesting solutions. We’re starting to look more closely at open source options. Apart from the big proprietary providers like Open AI, Microsoft or Google, we need to understand the open source options.

– I don’t think we should see AI as a machine that stands on its own. It must be deeply integrated into our systems and processes. This is why all our system partners must use AI components in their environment. It should find its way across the entire computing landscape, and it will.