Carmakers adopt AI agents faster than the rules for them

The Capgemini Research Institute published a study on 12 October based on a July survey of 800 senior executives at carmakers, tier-one suppliers and technology providers in 12 countries. 68% already use (33%) or pilot (35%) agentic AI in product development, but only 5% describe their AI governance as fully mature.

• Ownership is where it thins out. Only 10% have established shared accountability for the AI lifecycle across their ecosystem of partners, and 45% have not yet defined an AI strategy and roadmap for product development.
• Trust falls as the work moves closer to the vehicle. Executive trust in agentic AI stands at 59% for upstream research and concept work and drops to 40% for development activities.
• Six in ten executives say delays come mainly from process inefficiency rather than from the limits of the technology. Among carmakers, 46% in Europe and 53% in the US are implementing AI initiatives.

In an industry where one vehicle programme runs through a long chain of suppliers, the 10% figure matters more to us than the 68%. An agent that drafts a specification or a test plan inside one company is a productivity tool. Once its output crosses into a supplier’s process, somebody has to own it, and in nine out of ten organisations that question is still open. Much of the supply base sits in Europe and in Germany in particular, and we expect suppliers to meet the governance question first as a customer requirement in their contracts, well before it reaches them as regulation. (Capgemini)