The challenge
Stellantis e-Transmissions is a leader in manufacturing advanced hybrid transmissions, creating systems that deliver both impressive driving performance and outstanding fuel efficiency for hybrid vehicles worldwide. Despite their technical expertise, the company had no internal AI specialists to leverage emerging technologies for testing and process optimization. This gap revealed itself in 2 concrete challenges:
First, their existing testing system to map product failures was costly and time-consuming. For example, every failure requires a full manual inspection, sometimes even forcing tests to stop for long periods. Although it’s an unavoidable part of the process, the statistical tools in place don’t offer enough contextual information about the failures themselves. This lack of insight makes it difficult to guide root-cause analysis and breakdown investigations efficiently, particularly in the case of critical breakdowns.
At the same time, Stellantis e-Transmissions, like many other companies, works with extensive documentation and data. Its usefulness, however, depends largely on employees being able to quickly find the right information at the right time. In practice, this places significant pressure on both the documentation infrastructure and the organization’s internal experts, who are often relied upon to bridge the gaps. To make the process less time-consuming and more efficient, Stellantis invested in a custom AI language model developed by an external agency. Nevertheless, the ongoing adoption and maintenance posed new challenges. Enter our AI and machine learning specialist, Harun Kalkanci!
Approach & solution
Harun jumped into the project headfirst. For the testing system, he developed a custom anomaly detection model that learned from healthy tests what ‘normal’ looked like. Once up and running, the model can flag unusual behavior, pinpoint which signals were off, and even measure how serious the anomaly was. Engineers can quickly review the results to separate real failures from harmless noise and decide whether a test should continue or stop. Over time, the insights from the data also allow them to finetune both testing and product development, making the process smarter and more efficient.
On the knowledge management front, the AI language model, previously maintained externally, was now fully brought in-house. As more teams adopted the tool, Harun refined it, adding new document sets, defining and setting up the right AI assistants per team, and rolling out new features like document and image uploads. Custom agents were key: without them, every query would have to sift through all documents, from HR to R&D, slowing response times and increasing costs. Now, the tool runs smoothly, efficiently, and is actively embraced across the company.
Impact & results
While the anomaly detection system for testing is not yet fully implemented, its potential is clear. It promises to dramatically reduce downtime by detecting failures earlier, filtering out noise and providing deeper insights into the real causes of breakdowns, rather than just the visually obvious ones. This will allow engineers to act faster, make more informed decisions and fine-tune both testing and development processes.
On the knowledge management side, the impact has already been tangible. Searching for documents has become faster and more efficient, colleagues with more experience and expertise are no longer a bottleneck, and the tool also serves as a resource for research and insight gathering. Conservatively, Stellantis e-Transmissions estimates that each search saves about five minutes. With over 5000 requests each month, that adds up to nearly 60 full working days saved per month—essentially the equivalent of 3 full-time employees’ effort.
Key takeaways
The case of Stellantis e-Transmissions demonstrates the value of having flexible, on-site expertise for complex, custom projects:
- Data-driven insights reduce downtime and cost: Implementing AI-powered tools like anomaly detection can cut testing delays, improve root-cause analysis, and prevent costly failures before they escalate.
- Scalable knowledge management transforms collaboration: Custom AI language models, maintained internally with the right setup, make information retrieval faster, more accurate, and reduce time spent searching or interrupting colleagues.
- Expertise accelerates results: Bringing in on-site AI and machine learning specialists can quickly identify opportunities for automation and smarter decision-making, even in technically sophisticated companies.
This story highlights how strategically placed on-site expertise accelerates innovation and ensures consistent, high-quality results. Want to learn more about how we can boost your R&D team? Find out here!