Talk to anyone who runs an automotive warranty operation and you'll hear the same story: A dealer logs a repair. Weeks later, someone at the original equipment manufacturer spots a pattern. Then the case moves to manufacturing, then to the supplier, then back to engineering.
Each stop is a separate system, with its own data and owner. By the time the loop closes, thousands more vehicles are on the road with the same problematic part.
That's the automotive value chain in miniature: a long line of handoffs. Every handoff between disconnected systems is a place where work can slow down or break.
AI agents can help, carrying a case from a dealer to engineering without waiting on a person at every stop. Most automotive companies already have the technology. They just haven't put it to work across the handoffs yet.
For our third annual Enterprise AI Maturity Index, ServiceNow and NVIDIA worked with ThoughtLab to survey 325 executives and 74 employees on automotive AI. Among the organizations represented, 66% are using agentic AI. Only 7% have made significant progress using it to build autonomous, multistep workflows.
Automotive firms are putting 14% of their IT budgets toward AI, up from 11% in 2025. Leadership is on board too. The question now is whether that commitment is turning into working systems. The survey points to a disconnect between deciding and doing.
We asked respondents to rate their progress in seven areas, from 0 (no progress) to 5 (fully implemented with significant results), and converted each to a score out of 100. Automotive firms scored 56.6 on vision and leadership, which covers setting a shared AI vision and choosing projects based on strategic outcomes.
Firms reported the least progress (with a score of 39) on AI-enabled workflows, which covers using AI to run work across functions, including autonomous, multistep workflows.
In other words, automotive firms seem to be further along in deciding what AI should do than in putting it to work. The problem isn't ambition or access to models; it's the systems underneath.
When work is split across disconnected tools and databases, AI has no clean, shared data to draw on. According to our research, 73% of automotive executives cite inadequate data accuracy, access, and management as a barrier to scaling AI. And 53% are concerned about integrating with legacy systems.
Volkswagen is working on the plumbing directly. Its 2025 annual report describes an effort to standardize IT infrastructure across its brands and regions so that systems and services can be shared rather than rebuilt for each one. The company treats data as the foundation for AI, noting that its measures for optimized data availability form "the basis for up-to-date AI applications."
Many automotive firms end up automating tasks: faster defect detection here, smarter call routing there. That's useful, but it can be built on top of fragmented systems rather than fixing them. The harder work is connecting data across the company so AI can operate across functions. Without that, a customer can encounter a company that behaves like several separate businesses.
Stellantis offers one example of how using AI agents to connect systems and workflows can pay off. The automaker put AI at the center when it unveiled FaSTLAne 2030, its five-year, 60-billion-euro ($67 billion) plan to refocus brands and simplify technology across a 14-brand group.
“We're looking at creating at least €6 billion [$6.74 billion] of value,” says Mo Fagir, global director of the enterprise platform and AI delivery team at Stellantis. “AI is a principal way to achieve that.”
The clearest early win came from the supplier portal. Previously, about 90,000 users had to navigate more than 100 applications and file more than 1,000 IT tickets per month just to find the right one, Fagir says.
Now, an AI agent fronts the portal. It handled more than 20,000 actions in its first week, and Stellantis estimates 50,000 hours saved. IT teams spend their time on the exceptions that need a person.
The foundation underneath is what made it work. A central AI office governs the program, with regional and functional leaders owning the portfolios in their scope.
"We manage and control the entire AI portfolio through the platform," Fagir says.
That's one view of every AI project across the company. It's also why supply chain, dealers, and purchasing are next, as systems inherited from years of acquisitions converge onto a single platform.
Stellantis fits the profile of what we call Pacesetters—the 17% of automotive firms scoring in the top tier of the Enterprise AI Maturity Index. They average 73 out of 100 on AI maturity. Other automotive firms in the study average 46.
The difference between the two groups is in how they build:
- 54% of Pacesetters integrate workflows across business functions, compared with 5% of other firms.
- 50% have replaced fragmented legacy systems with an integrated platform, versus 6% of others.
- 70% say they have processes to clean, tag, and standardize data, compared with 12% of others.
Pacesetters are also 9.6 times more likely than other firms to report productivity gains from AI. That has less to do with which model they use than how they connect data, workflows, and governance so AI can finish the work instead of handing it back to a person at every step.
In the automotive industry, safety and compliance aren't optional. That’s why governance can't show up after deployment.
According to our research, 71% of Pacesetters have created an AI Center of Excellence—a team that provides guidance, standards, and support to ensure AI is used responsibly—compared with 14% of others. That's what lets governance keep pace with deployment instead of arriving after it.
"Agents are powerful, but especially in the automotive industry and manufacturing, they need context, guardrails, and a secure environment along with human oversight,” says Sri Subramanian, global head of generative AI for automotive at NVIDIA. “Workflow automation and orchestration go hand in hand with governance."¹
In practice, that means a control tower that audits what AI agents do, enforces guardrails, maintains a compliance trail, and keeps a person accountable for high-stakes decisions.
Most automotive firms we surveyed (62%) are Scalers, meaning they're moving from plans to implementation but still working through siloed systems. For these firms, 100 experiments across departments matter less than getting two or three high-impact workflows into production.
We recommend picking one that crosses functions. Warranty and recall management are strong candidates because they touch the dealer, the original equipment manufacturer, manufacturing, suppliers, and engineering.
Map every system the workflow touches, fix data ownership along that path, and then let AI agents run it end to end, with humans stepping in where judgment is needed.
A recurring defect that today takes three years to investigate could be surfaced in three months. That's what it can look like when the value chain moves as one.
Take our seven-minute Automotive AI Maturity Assessment to see where your organization ranks. And get more insights in the full Enterprise AI Maturity Index for Automotive.
