What ChatGPT can do for automotive operations

Munich, May 2023

What ChatGPT can do for automotive operations

Munich, May 2023

he AI tool everyone is talking about can code websites, write articles and pass exams. But what does the technology offer to the operations teams of automotive OEMs and suppliers?

“ChatGPT is a large language model developed by OpenAI, capable of generating human-like responses to natural language input. It is trained on a vast amount of data and can be used for a variety of applications, such as chatbots, language translation, and text completion.”​

This is how ChatGPT describes itself when you ask the artificial intelligence (AI) chatbot “what is ChatGPT?” It’s pretty accurate – although the description covers only a limited part of the tool’s range of uses. It is also out of date – it doesn’t mention, for example, that the latest version, which uses the GPT-4 language-processing model, can also extract information from images to produce answers.

Without a doubt, ChatGPT exemplifies the potential that AI has to revolutionize the way we live and work. But what specifically does this type of generative AI has to offer the operations teams of automotive OEMs?

Our view is that there are at least two big advantages: firstly, the language model can be trained with huge amounts of information, identify trends invisible to the human eye, and can “interpret“ the results, to make recommendations or draw conclusions on its own.

Secondly, it has an easy-to-use interface because it deals with natural language. ​It can therefore be used by engineers and operations specialists with no background in AI to make informed decisions in real-time. In addition, the ability of newer versions to analyze information in the form of images as well as text, offers limitless potential in automotive operations.

In this article we will look at six potential Operations use cases for generative AI tools. However, it is also important to sound a note of caution about the risks posed by using machine learning algorithms. Like all forms of artificial intelligence, the quality of the information provided, or decisions made, relies on the quality of the data it is trained on. And even if every single piece of training data is correct, the AI could still come to the wrong conclusions because of the way it combines them. Currently ChatGPT, for example, does not highlight different levels of assurance in the results it delivers.

Complex production issues will therefore still require human expertise and intuition. Knowing the cases in which such human expertise is required is unfortunately not easy, as it needs precisely the subject matter expertise that a user of ChatGPT may not have.

Another disadvantage is the potential security risk – we would not recommend integrating sensitive product and production data with generative AI on a public cloud service.

As a result, OEMs and suppliers must ensure that employees are aware of the risks involved in leveraging the technology, and have the appropriate data classifications, security measures and usage rules in place to protect confidential information.

However, we assume that technical solutions will become available to address these challenges, and so it makes sense for operations teams to start looking at potential use cases now. We have identified seven areas of operations where generative AI could offer significant benefits:


1. Concept phase: how can we learn more about what the customer thinks?

OEMs spend millions learning about how customers perceive a car and its features. In the development stage, potential customers are brought in to carry out interior design reviews and test drives. Nevertheless, errors often seem to fall through the cracks. Think of the mockery German car manufacturers had to endure when they first introduced cupholders in their cars for the US market: they were much too small and the wrong shape for customers there.

A generative AI tool could be used to assess large volumes of customer test drive and review responses, spoken or texted directly into an app. A very accurate survey of customer sentiment could be generated based on this. A developer could then feed their latest drawings into the software and ask the generative AI whether the design would be appreciated by customers, based on its assessment of the responses.

2. Requirements engineering: can generative AI streamline supplier interactions?

When OEMs buy parts from suppliers, they start by transferring a list of requirements. In practice, the requirements are often taken from the previous version of the module or system, and adapted for the new order. This means that in most cases, that number of requirements increases from one generation of cars to the next, as engineers add more requirements but rarely delete any of the old ones.

The sheer number of requirements, and the limited relevance of some of the older ones, leads to situations where suppliers don’t actually read and understand the full list before they start developing a product. For a part such as a battery inverter, a generative AI application could be used to crosscheck which requirements actually apply to the specifications for the new version, and to highlight the legacy items which can be cut. Both OEMs and suppliers would benefit from more precise requirements and more focus in the development stage.

3. Quality improvement: how can we really access lessons learned?

All automotive OEMs and most suppliers have databases in which they document faults and problems over the lifecycle of a product, as well as the measures taken to solve those issues.

These databases contain tens of thousands of line items, with lessons learned from many engineering projects. However, this valuable knowledge is buried in not very user-friendly programs and cumbersome to use. The result is that during both the development phase and series production, engineers often don’t have the right information to hand, despite the fact it is held by the OEM. This leads to avoidable design faults, and fault resolution during series production takes longer than necessary.

Generative AI applications could be a user-friendly interface between the engineer and those vast databases. It would be hugely valuable for OEMs if AI could quickly find the “right“ lessons learned and provide them to the engineer. However, this is a prime example of a task where keeping a “human in the loop” is essential, to make sure the right questions are being asked and that the results coming back are addressing the specific quality issue. 

4. Maintenance: how can automation reduce stock levels and speed up repairs?

Generative AI can compare the specifications of parts against existing in-stock parts or search for alternative spares, to prevent downtime in out-of-stock situations or find cheaper solutions. In addition, AI could be used to quickly analyze maintenance and past damage reports to identify the most frequently ordered spares that need to be in stock at all times.

When it comes to repairs, we see a number of ways that such an AI tool could be useful. Its language processing capability could be used to filter through past damage reports, manuals, and the history of a part to quickly find the right solution for replacing and maintaining it on the shop floor, as well as in the factory.

One of the main potential uses for ChatGPT specifically is as a chatbot, and it could be used for troubleshooting during a repair, comparing the machine manual to damage reports and documented repair procedures, to identify promising solutions and minimize downtime. In the near future, AI chatbots might also be capable of providing first and second level support for customer and service personnel during a repair.

5. OEE: how can generative AI increase factory productivity and profitability?

AI could be used to analyze shift handbooks, maintenance logs and quality reports and link the information to certain events, such as shut downs, quality problems and setup-times. By working in combination with a data analytics solution that identifies patterns in process parameters, a generative AI tool could identify and suggest optimal set-ups and schedule preventive maintenance to prevent potential equipment failures.

Armed with this information, production managers can make more informed decisions and take corrective actions to improve OEE, and increase factory productivity and profitability as a result.

6. Supply chain: where can AI cut the burden of paperwork?

Generative AI could be used in supply chain operations to analyze and interpret the complex legal texts linked to deliveries, including tariff laws and regulations from different countries. These are subject to frequent changes that can be challenging for supply chain managers to keep up with. By processing vast amounts of legal data quickly, generative AI can identify the relevant new changes in legislation and provide actionable insights.

The AI could also be trained to identify opportunities for companies to leverage differences in tariff laws and regulations from country to country, in order to reduce tax costs and speed up the operation of their supply chain. Both examples will help supply chain managers to design more efficient and cost-effective networks.

7. On the road: how can generative AI lead to significantly higher customer satisfaction?

Despite all the automotive industry’s quality management processes, there are still occasionally problems with cars once they are on the road. We have written in the past about the potential for OEMs to make better use of field data to catch and address such issues early, and with the advances in AI, potential solutions to help them do so have moved closer. Today, a customer with a problem while driving phones the OEM’s call center. But with a chatbot powered by generative AI, instead of working through the possible reasons for failure together on the phone, the customer support worker could ask the chatbot what the most likely problem was, feeding in the customer’s report of the fault and the car’s over-the-air-failure codes.

One of the next steps in connected vehicle development is likely to be to integrate AI functionality into the car itself so the driver can ask for help directly via voice commands. For example: Driver: “Why is the rear camera not working?” Car: “The camera is not working because the trunk is open.“

As automotive operations experts, we are thrilled to consider these possibilities and to work with clients to assess the questions and bring solutions to life. Yet undoubtedly, the impact of using generative AI tools in automotive operations extends beyond the factory floor and into society as a whole. As with all technologies, vast language processing capabilities can be abused as well as used for their intended purpose.

Concerns about data privacy and security remain, and as the technology becomes mainstream, it is almost certain to replace jobs. Developers, quality engineers, production planners and maintenance workers – every role that is dependent on experience and gathering knowledge to find new solutions could be at risk. ​However, while the AI remains generative – fed with data in order to produce outputs and make decisions, rather than having original thoughts – there will be still be a need for human experience and expertise alongside.

And what will the future bring ? Generative AI is here to stay. That is certain for us. It’s hard to predict what specific applications will emerge from the new tools that appear almost daily. They will significantly transform life on the shopfloor and development departments. The efficiency of indirect processes will increase drastically. Ultimately, vehicles and production processes will be completely developed by AI. And probably more flawlessly than by humans. We will stay tuned.

Heiko Weber


Fritz Metzger


Christian Grimmelt


Timo Kronen


Philipp Stütz

Associate Partner

Bolko von Hochberg


Heiko Weber

Heiko Weber (1972), Partner at Berylls Strategy Advisors, is an automotive expert in operations.

He started his career at the former DaimlerChrysler AG, where he worked for seven years and was most recently responsible for quality assurance and production of an engine line. Since moving to Management Engineers in 2006, he has been contributing his experience and expertise to projects for automotive manufacturers as well as suppliers in development, purchasing, production and supply chain. Heiko Weber has extensive experience in the development of functional strategies in these areas and also possesses the operational management expertise to promptly catch critical situations in the supply chain through task force operations or to prevent them from occurring in the first place.

As a partner of Management Engineers, he accompanied the firm’s integration first into Booz & Co. and later into PwC Strategy&, where he was most recently responsible for the European automotive business until 2020.

Weber holds a degree in industrial engineering from the Technical University of Berlin and completed semesters abroad at Dublin City University in Marketing and Languages.

Fritz Metzger

Fritz Metzger (1986) joined Berylls Strategy Advisors, an international strategy consultancy specializing in the automotive industry, in February 2021. He is an expert on automotive operations.

Since 2011, his focus has been on strategic alignment and operational efficiency improvement of automotive manufacturers and suppliers. He also advises top management in critical situations, including R&D and industrialization task forces and relocation and restructuring initiatives of plants and complete suppliers. The challenges of e-mobility are always in focus.

Before joining Berylls, he was a director at international strategy consultants PwC Strategy&, as well as a sales and project manager at a medium-sized supplier and mechanical engineering company.

Fritz Metzger is a trained industrial engineer with a degree from ESB Business School Reutlingen. He also holds an MBA from the University of Salzburg.

Christian Grimmelt

Christian Grimmelt has been an integral member of the Berylls Strategy Advisors team since February 2021. Previously, he gained extensive professional experience in top management consultancies and in the automotive supplier industry.

During his time at the world’s largest automotive supplier, he drove the establishment of a central unit to optimize the company’s global logistics and production network.

Christian Grimmelt’s consulting focus is logistics and production network optimization, purchasing and (digital) operations including launch and turnaround management for OEMs and especially suppliers.

Christian Grimmelt holds a university diploma in industrial engineering from the Karlsruhe Institute of Technology.

Timo Kronen

Timo Kronen (1979) is partner at Berylls Group with focus on operations. He brings 17 years of industry and consulting experience in the automotive industry. His focus is on production, development and purchasing as well as supplier task forces. Some of his recent projects include:

• Restructuring of the Procurement Function (German Sports Car OEM)

• Supplier Task Force for an Onboard Charger (German Premium OEM)

• Strategy Development for the Component Production (German Premium OEM)

Before joining Berylls, Timo Kronen worked at PwC Strategy&, Porsche Consulting Group and Dr. Ing. h.c. F. Porsche AG. He holds a diploma degree in industrial engineering from the Karlsruhe Institute of Technology (KIT).

Philipp M. Stütz

Philipp M. Stuetz (1981) joined Berylls at the beginning of 2021. He has over fifteen years of experience in the automotive industry. Thereof he spent seven years at an international automotive supplier with assignments in Spain, the USA and Mexico and over eight years in consulting. His focus is in operations excellence, especially in large transformation programs, process optimizations and efficiency improvements in administrative functions and indirect operations areas. He counts suppliers and OEMs to his clients alike.

Philipp M. Stuetz graduated in business administration from the universities of Stuttgart and Strasbourg.