Three momentous trends are buffeting the R&D function in the automotive industry, creating the need for profound change.
First, the transition from internal combustion engine (ICE) to electric vehicle (EV) technology is a fundamental shift, the likes of which the industry has not experienced since surging oil prices and competition sparked demand for more fuel-efficient vehicles more than a half century ago.
Second is the trend of software-defined vehicles with a central architecture that is more geared toward consumers. Software provides many opportunities for automotive players to differentiate themselves, with such applications as infotainment and advanced driver-assistance systems. However, software also presents companies with the substantial challenge of transforming hardware-centric operations to support their added role as software providers.
The third trend is the emergence of generative AI (gen AI). Gen AI is becoming a powerful technology with the potential to completely reconfigure how R&D teams operate. Although the technology is still in its early days, its ability to generate and process language and imagery, integrate insights from various sources, process information across diverse formats, and produce detailed documentation for regulatory purposes points to a radically different R&D future.
New entrants to the sector—EV manufacturers in China, the United States, and elsewhere—have already successfully implemented R&D process innovations that accelerate new-vehicle time to market, gaining considerable strategic advantages over established players, whose margins are already squeezed.
To better understand the impact and opportunity of these trends, we spoke with executives from leading European automotive and manufacturing companies. The detailed discussions focused primarily on gen AI and the lessons that are emerging from the many gen AI pilots and a few at-scale deployments.
One clear lesson emerged from these discussions: by following a value-focused approach that supports the integration of gen AI throughout the R&D process, companies can capture substantial value in the form of reduced costs, accelerated time to market, improved quality, and more innovation.
Opportunities for gen AI in automotive R&D
We convened a workshop with 30 R&D executives from leading European automotive and manufacturing companies to discuss their use of and plans for gen AI, exploring a range of opportunities inherent in the technology. Some of these executives also completed a detailed survey on gen AI; their responses are reflected throughout this article.1
Adoption and investment trends
We found a strong inclination to adopt gen AI in the automotive sector. The majority of companies (75 percent of survey respondents) are experimenting with at least one gen AI application; those that are not plan to start within one year (25 percent of respondents).
Further, investments in gen AI applications for R&D are substantial: more than 40 percent of survey respondents reported that their companies have invested up to €5 million. A few companies—more than 10 percent of respondents—have invested more than €20 million.
Gen AI application in R&D processes
While most of the executives surveyed (70 percent) reported that their companies are integrating gen AI applications into R&D, most pilot programs are limited to one stage of the R&D process. The breadth of piloted use cases is remarkably high (60 percent of use cases we identified); nevertheless, we saw no systematic use of gen AI throughout the R&D process.
The wide range of piloted use cases indicates that executives are largely aiming for a comprehensive future approach to using gen AI in the R&D process. Indeed, more than 40 percent of survey respondents are prioritizing more than 75 percent of potential use cases.
Gen AI’s estimated impact and value
Most participants agreed that most of the gen AI use cases quantified in the survey or during the workshop bring substantial value and could improve R&D processes by 10 to 20 percent. Some participants viewed the potential value of integrating gen AI use cases more as a means of recovering investments in gen AI, while others viewed it as an added expense required to remain competitive with peers. Nonetheless, the prevailing view is that major organizational and cultural transformations are needed to capture gen AI’s full value.
Gen AI use cases and their potential value to R&D
Use cases currently being piloted or investigated by participants’ companies were most likely to focus on requirements engineering (mentioned most in the survey), followed by software testing and validation, and product design and optimization (mentioned by more than half of survey respondents).
Although these are the most frequent areas of focus, each segment of the R&D process has viable gen AI use cases that provide opportunities to reduce costs, increase speed to market, and improve quality. For example, administrative costs could be lowered by using gen AI to complete certain documentation tasks required by regulations, thus freeing up developers’ capacity and improving engineering experience and efficiency.
- Testing and homologation. The executives we consulted estimated that using gen AI to automate reporting and to generate documentation and scenario-based simulation could improve testing and homologation processes by 20 to 30 percent. Automation could add value by simplifying the creation of essential reports, manuals, and documentation for compliance, product documentation, and quality assurance purposes.
And some use cases can deliver exceptional efficiencies: for example, a German tier-one automotive supplier achieved a 70 percent gain in productivity—including the time required for human review of the gen AI–produced output—byusing gen AI to generate test vectors such as full branch coverage and modified condition/decision coverage (MCDC). By integrating gen AI into its development process for embedded software and its generation of requirements—using gen AI to help determine requirements for stakeholder requests that could serve as first drafts—the company achieved productivity gains of up to 30 percent for engineers. - Design applications. Within the design segment of R&D, the leaders we consulted estimated that generative-design use cases could improve R&D processes by 10 to 20 percent. They also estimated that reverse and black-box engineering use cases could yield 5 to 10 percent improvements in R&D processes by using gen AI to reveal and decode proprietary technologies such as knowledge extraction, algorithm decoding, or reengineering.
Capturing gen AI opportunities
Most of the executives we consulted deemed the barriers to implementing gen AI in R&D as either “large” or “very large”; only 25 percent of survey respondents characterized them as “small.” Indeed, the lack of systematic integration of gen AI into companies’ existing operating models can be attributed to the major organizational and cultural changes such integration requires.
For gen AI applications to add value across the R&D process, a holistic, value-centered approach that goes beyond tech and data is needed. Only by building the range of necessary capabilities and culture can companies expect to reap the benefits of new technologies such as gen AI (exhibit).
A road map centered on value
A surprising number of transformations lack clear and specific targets for value. Without that clarity and alignment at the leadership level, companies struggle to marshal the necessary resources and track progress. Building support and alignment around that value is critical.
- Frame gen AI as an enabler. A major theme in our discussion with R&D executives was avoiding gen AI backlash in the organization by properly positioning the benefits of gen AI via a preemptive discussion. Presenting gen AI as an enabler and accelerator rather than as a means of cost reduction and job destruction is critical to a successful adoption.
- A clear and consistent change narrative. Internal stakeholders—chief experience officers, managers, employees, and relevant departments such as legal, ethics, and compliance—should be engaged in the process of defining the change narrative. This collaborative approach ensures that all perspectives within the organization are considered and that the resulting narrative is comprehensive and aligned with the organization’s goals. The change narrative should address ethical considerations, including data privacy, algorithmic bias, transparency, and accountability, and it should be consistently communicated to all stakeholders (see sidebar, “Addressing legal and ethical considerations”). This helps build trust, understanding, and support for gen AI implementation and encourages all members of the organization to support its strategic objectives.
- Empower C-suite leaders. A crucial first step can be providing C-suite leaders with relevant data and case studies that demonstrate the potential impact of gen AI on the organization’s strategic objectives in a clear and concise manner. In an ideal situation, C-suite leaders are briefed on the ethical and legal considerations of gen AI and the importance of establishing guardrails and nimble ethics and legal approval processes. By modeling a pioneering mindset, C-suite leaders can foster a culture of innovation and experimentation within the organization. R&D departments can support these leaders in ensuring that gen AI is implemented in a responsible and effective manner that maximizes benefits for the organization.
- Build a visible lighthouse to inspire the organization. This can prove to be a key strategy for R&D departments in automotive. By identifying a high-impact use case for gen AI and showcasing its potential, R&D teams can motivate and inspire the rest of the organization to explore new possibilities and embrace innovation. Once the lighthouse is successful, R&D leadership will need to build a set of mutually reinforcing and supporting use cases. Simply building out use cases in isolation and without coordination often leads to considerable activity but little value.
Empowering and training talent: Two gen AI copilots
Gen AI will certainly affect jobs, but McKinsey analysis indicates that models will often function as copilot applications that support the work employees are already performing.2 The changes will include assuming monotonous tasks, such as writing briefs or drafting documentation. This in turn will enable employees to spend more time on more rewarding tasks—for example, generating ideas and creative solutions and drafting initial code for review. Building talent capabilities can be much more of a training exercise than a hiring objective.
Several R&D organizations have already begun implementing copilots, including one focused on writing requirement documents. Copying and pasting text from different versions of such documents often leads to inconsistencies and, in the worst cases, convoluted documents with unnecessary requirements. A German OEM that implemented this copilot has realized efficiency gains of 20 percent and eased the workloads of several hundred engineers. The copilot has been continuously enhanced with additional features, such as checks against International Organization for Standardization (ISO) norms, that have contributed additional time savings.
Another gen AI copilot, implemented by a German automotive OEM, aims to reduce compliance task preparation time for a broad set of employees. In this case, the gen AI application automatically extracts norms from ISO and similar documents, consolidates them, and checks for adherence to and compliance with existing process documentation. It is expected to reduce audit preparation efforts by 20 to 30 percent when it is eventually expanded to derive to-do items automatically and identify synergies across norms and process documentation.
Workers will need time and training to learn how to best use their copilots, test and build trust in the results, and gain reassurance from interactions that yield the correct answers.
Innovating the operating model
For teams to work quickly and effectively, they need independence, clear guidelines and goals, and access to gen AI tools and capabilities. Given the uncertainties and evolving landscape around risk, experts in the field need to be embedded with working teams to identify issues early on and manage a thoughtful risk review and approval process. Crucial elements include the following:
- Cross-functional teams. To fully leverage gen AI, companies should establish cross-functional teams consisting of experts from various disciplines who can collaborate effectively to drive innovation. A culture of collaboration and experimentation enables teams to solve complex problems and deliver powerful solutions.
- Streamlined processes. To unlock the bottom-line potential of gen AI, existing processes must be redesigned and costs must be systematically reduced or removed—for example, by streamlining workflows, eliminating manual process steps, and reshaping roles.
- Clear mandates. To ensure accountability and drive results, leaders should establish robust mandates that clearly define objectives, deliverables, and timelines for these teams. By empowering teams with the necessary resources and authority, R&D departments can foster a sense of ownership and responsibility among team members, enabling them to achieve their goals and deliver tangible outcomes.
Building technology foundations
Similar to digital use cases, the top barriers to adoption of gen AI use cases include data silos, permission issues, and tech stacks that prove inadequate to support the new technology.
Implementing gen AI depends on scalable infrastructure, which includes robust architecture, efficient resource allocation, and proactive adaptation to evolving technological landscapes. In addition, a coherent but modular data platform is an essential element to a technical foundation that supports a scalable use of gen AI. Ideally, the technology foundation should provide access to different gen AI models to enable broader sets of use cases and support cost-efficient implementation. The vast amount of data to be processed and the open architecture required to integrate with vendor-hosted large language models mean that cloud-based infrastructures and platforms are desirable, as they can provide the flexibility and robustness needed.
Creating robust data governance
A lack of value assurance for data and a lack of training-data availability from suppliers are substantial challenges, highlighting the technical and organizational centrality of data.
Although gen AI does not typically require vast amounts of data to deliver value, most use cases do benefit from systematic prompt enrichment using proprietary data, which needs to be administered with strict data governance to restrict visibility and accessibility of the proprietary data to what is permissible and desirable.
Data ownership, data taxonomy, and ontology are required to feed models with the necessary clean and representative data for training. A comprehensive approach is not ideal; rather, a set of pragmatic solutions for various use cases, introduced in parallel, is more likely to yield success. For example, companies can begin cataloging test cases, establish data governance for these test cases, and create a repository of high-quality test-case data in a structured data lake.
Maintaining strict assurance that solutions are adopted and scaled to capture value
The effects of the transformed R&D process should be carefully measured, evaluated, and corrected where needed. A clear and accepted baseline is essential to demonstrate the positive impact of gen AI.
Similarly, a value capture governance framework and supporting incentives will help avoid unnecessary costs for licenses and training. An organization with a strong business value can help realize bottom-line or top-line benefits enabled by higher productivity and shorter time to market.
R&D departments can be better served by prioritizing gen AI use cases with the highest potential impact and lowest risk. These use cases can then be grouped into waves of deployment based on their complexity and interdependencies, thus delivering tangible business value and building momentum for the next wave of deployments. Collaboration and knowledge sharing across teams and waves can help maximize benefits.
Capturing the value of gen AI in R&D starts with a clear vision for its use and a systematic approach to identifying and prioritizing use cases. Each use case pilot should be followed by developing a product that is supported with robust change management, value accrual, capability building, and a road map outlining the next wave of use cases to transform the entire process chain. Value can be captured only by ensuring that gen AI benefits are applied day to day.
Implementing several use cases, partly in parallel, facilitates successive iteration and refinement of the gen AI approach, strategy, and vision over time. Such an approach ensures a swift and scalable capture of value from gen AI innovations.
