Stellantis Engineers Use AI to Cut Development Time

Published with our partners' assistance

Stellantis is pushing AI into its engineering to shrink vehicle development from four years to 24 months, with two models due in 2027. Engineers at its Michigan technology center say yearlong projects now take weeks.

A thin margin behind the push

Stellantis runs 14 brands, and its European arm is losing money. Its adjusted operating margin was 1.8% in the second quarter. CEO Antonio Filosa wants 7% by 2030.

Of the three Detroit carmakers, Stellantis depends most on Europe, which supplied about 38% of its 2025 revenue. It also bet heavily on EVs in North America. After the $7,500 federal EV tax credit ended on Sept. 30, 2025, it took the biggest hit in Detroit.

In February the company wrote down roughly €22.2 billion, about $26.5 billion. Shares fell 25% that day and are down nearly 60% year to date.

Mel Roberts, director of North American infotainment, plans features five to eight years ahead. Her biggest surprise of recent years was not AI. It was the collapse of electrification ambitions, which she says reached into all of the company's future plans.

Weeks, minutes and a caveat

Joydip Saha runs the hands-free driving project. His team finished the system in a Grand Cherokee test vehicle in six weeks using AI tools. Comparable work used to take a year or 18 months.

He bars AI from system architecture, because people would lose the ability to debug. Safety-critical work, he says, demands engineers who can answer every question.

Arturo Guzman's aerodynamics team brought AI in about five years ago. He says there is no magic in it and compares the tool to an intern. A front splitter change on a Ram pickup now gets a drag estimate from his team's model in about three minutes. Before, the full simulation ran 14 to 24 hours.

Ben Zawodni, a software engineering manager, is folding hundreds of electronic control units into a central computer called STLA Brain. His team's first working version of a company-wide observability platform, a tool that watches the software deep within the car's computers, took a week. He recalls about six months before agentic coding tools. Fixes that once took two to three weeks now finish in two to three days.

Not everything has sped up. The wind tunnel usually tests one car a day, according to its supervisor Steve Kubiak. Roberts says her AI research tool returns 10 to 15 sources and she checks every one. She does not think it saves time, though her work becomes more robust.

Almost every engineer who reported faster work also said they verify what the AI produces.

Alison Rahm leads vehicle engineering for midsize and large SUVs. In her telling, AI now handles review work once done manually, such as checking a design against roughly 1,000 engineering practices. She still double-checks. She notes that even once crashes could be simulated, engineers kept crashing real cars into walls to confirm their assumptions. For AI she puts the time to full trust at four years, which would land in 2030, the year of Filosa's 7% margin target.

Curic's pitch: make yourself redundant

Ned Curic, Chief Engineering and Technology Officer, came from Amazon, after earlier jobs at Microsoft and Toyota. He says the organization as a whole is not where it needs to be, even though some teams show strong results.

From Amazon and Toyota he took small teams and one owner per system, so engineers cannot blame each other when something breaks.

His hardest sell is the idea that “the best thing that you can do for your career is to make yourself redundant.” It works only if the company always has a new opportunity for those people, he says, and leaders need maturity and confidence.

Curic is also hiring. At the Paris auto show in October he told Bloomberg that Stellantis would add several hundred engineers in Europe and the U.S. this year.

Zawodni describes his team's reaction as mixed. “Engineers are really afraid of losing their core skills,” he says. He wants them to grow into architects. The tools have not eased his own load: he estimates he works twice as hard, at least, compared with before.

Where the research bets go

Anne Laliron leads a tech research group of roughly 400 people. Laliron says the company has to choose its bets: battery cells are designed in-house, and startups cover other technologies. Wayve, a British startup, supplies perception software for hands-free driving. Tiamat works on sodium-ion batteries and Zeta Energy on lithium-sulfur. Stellantis also holds a stake in the Chinese EV maker Leapmotor.

The largest bet is a battery-electric Dodge Charger prototype with a solid-state battery from Factorial, a Massachusetts startup that has been Stellantis's partner since 2021. Factorial co-founder Siyu Huang puts the technology three to five years from automotive production. Laliron drops about 10% of her group's projects each year when they fail.

Our View

The speed gains are reported per task, while the problem sits in margin points. A 1.8% margin against a 7% target is unlikely to close on faster simulations alone. The first test is whether the 24-month schedule holds for the two 2027 models. Another is whether teams like Zawodni's actually move into architecture roles.

Source: Fortune