Nearly seven in ten automotive organisations are already using or piloting agentic artificial intelligence (AI) for product development, but only 5% report having fully mature AI governance, according to a new report by Capgemini Research Institute.
The report, Rewiring Automotive Product Development: How Organisations Can Pull Ahead in the AI Race, found that 33% of organisations were using agentic AI, while another 35% were running pilot projects. Adoption in non-core engineering activities is expected to rise from 26% currently to 68% over the next two to three years.
However, gaps in data quality, traceability, accountability and collaboration across the automotive ecosystem remain obstacles to wider deployment. More than eight in ten executives surveyed believe agentic AI will allow engineers to spend more time on core engineering tasks by reducing administrative and coordination work.
The report, based on a survey of 800 senior automotive executives across 12 countries in North America, Europe and Asia-Pacific, also found significant differences in the business returns from AI. Organisations generating the most value reported development cycles that were 17% faster, engineering productivity gains of 19% and a 19% reduction in time-to-market.
According to Capgemini, these organisations have stronger data foundations, clearer business cases and more structured governance and operating models, suggesting that AI adoption alone does not guarantee better results.
“Agentic AI is creating a new performance divide in the automotive industry,” said Laurence Noël, Global Head of Automotive at Capgemini. She said organisations generating the most value were redesigning workflows, strengthening collaboration across their ecosystems and establishing clear accountability for AI-enabled decisions.
The report also identified trust as a constraint on the use of agentic AI in engineering. While executive confidence in applying the technology to upstream research and concept activities stood at 59%, it fell to 40% for development activities, where AI-generated outputs can have more direct implications for engineering outcomes, product quality and safety.
The findings also point to limitations in how automakers use data generated by vehicles already on the road. Some 64% of organisations said insights from vehicles in use primarily inform future vehicle programmes rather than improvements to vehicles already on the road. Over-the-air updates remain concentrated on areas such as cybersecurity and navigation, with fewer organisations using them to modify driving functions or assisted-driving capabilities.
Organisational challenges are also slowing AI deployment. Six in ten executives said delays were driven primarily by process inefficiencies rather than technology limitations, while 45% of organisations had yet to define an AI strategy and roadmap for product development.
Collaboration across automakers, suppliers and technology partners remains another gap. Only 30% of organisations reported having co-development models based on shared outcome metrics, while 10% had established shared lifecycle accountability across the ecosystem.
The report suggests that scaling agentic AI in automotive product development will depend not only on technology adoption, but also on governance, trusted data, clearer accountability and closer coordination across the vehicle development lifecycle.