Continental relies on Microsoft Azure AI: A game changer in R&D requirements management

The challenge: Highly manual and time-consuming processes in requirements analysis

Continental develops pioneering technologies and services for the sustainable and connected mobility of people and their goods. Founded in 1871, the technology company also develops intelligent driver assistance systems, connected cockpit technologies, and pioneering software integration. A structured and efficient development process is the essential factor here.

It may stretch across the entire width of a car’s cockpit, but even the curved pillar-to-pillar display started out as a requirements document. Automotive manufacturers use these documents to record the specific requirements for a new display, broken down into detailed functional and non-functional requirements. The documents can run to hundreds of pages—and feature up to 30,000 individual requirements.

In the past, the requirements engineers at Continental Automotive would analyze such documents manually. The first step was to review the document and gain an initial overview. In the subsequent detailed analysis, engineers would look at each requirement, categorize it as functional or non-functional either for the system or for its components, and assess its relevance to aspects such as safety or security. They would also check the requirements for consistency to ensure that there were no contradictory or duplicate requirements. Components that were not relevant for the development team at Continental Automotive were marked as such in consultation with the respective vehicle manufacturer. 

Finally, the requirements engineers would draw up a structured report aggregating the requirements they had analyzed and clustered. Preparing that kind of report was one thing above all: time-consuming. “We estimate that each individual requirement takes 140 minutes to go from analysis to implementation. Or around 10 minutes if it is a requirement that is not relevant for our development team,” calculates Michael Sicker, Director R&D Transformation at Continental. “In total, that is over 30,000 hours per project.” A substantial time investment, but one that is crucial for the progress and success of any project. The development phase cannot begin until the technical requirements have been correctly understood and agreed on with the experts involved.

Continental recognized the urgent need to reduce the manual effort involved in requirements analysis in order to increase the quality of the results, while simultaneously making more efficient use of development resources. “Our developers have lots of visionary ideas, and generative AI gives us an ideal opportunity to turn these into reality,” Sicker says. Together with Microsoft and NTT DATA, Continental Automotive developed a solution based on Microsoft Azure AI services and the latest advances in generative AI. This technology makes it possible not only to analyze requirements automatically but also to understand them in context, minimize errors, and accelerate end-to-end processes in the development cycle. 

“Microsoft and NTT DATA were more than just technology partners for us—they shared our vision and helped us take this solution to the next level,” Sicker says. “The result is an innovative application of generative AI that shows how AI-based innovations can transform the automotive industry.”

The solution: AI-based requirements engineering

Today, the process begins with “AI-based requirements engineering”: an automated review of the entire requirements document. The AI-based solution scans the document and identifies relevant sections and keywords within a very short time. To achieve this, Continental’s Automotive group sector uses prompt engineering, which involves giving the AI precisely formulated questions or instructions, called prompts, with the aim of obtaining the desired data. The AI models are constantly being improved through iterative refinements to the language, context, and structure of the prompts.

This solution provides employees with a quick overview of the content and structure of the requirements document without their having to go through it manually. “What we are talking about here is the potential for AI-based requirements engineering to greatly reduce the time aspect in particular,” Sicker says. In addition to the automatic review and categorization of each component, the solution compares the requirements with a catalog of generic Continental Automotive features. The User Experience division at Continental’s Automotive group sector has developed dozens of these generic features, which can be used in various projects with minimal customization.

Following on from architecture and design, the product goes through various test phases: module tests are followed by integration tests, system tests, and finally acceptance tests. “By drastically reducing the effort that has to go into requirements analysis, we can move on to product development and prototyping at an earlier stage. This lets us identify and solve critical problems long before the project is completed—a game changer for quality, efficiency, and successful product launches,” Sicker says.

“Basing the solution on state-of-the-art Azure AI technologies proved to be crucial for the success of the proof of concept within just three months,” says Jens Krüger, Head of Global Automotive Engineering at NTT DATA. Continental Automotive plans to scale the solution within the company and use it in other projects in various departments to further increase the efficiency and quality of its development processes. The company is constantly working to improve and adapt its AI models so it can meet ever-changing requirements and challenges.