GM works to snuff out glitches as it eyes big software profits

DETROIT — General Motors has placed a giant bet on software, investing billions of dollars that it believes will soon pay off big.

The automaker anticipates that software and services could generate as much as $25 billion in revenue by the end of the decade. With software becoming a key pillar of its business strategy, the company is working to line up the people and processes necessary to move faster and improve execution.

New executives with experience at tech giants Apple and Microsoft oversee a team that CEO Mary Barra has said is now "several thousand people" strong. It has recruited software talent from Silicon Valley to work on advanced driver-assistance technology, infotainment systems and in-vehicle connectivity.

However, the build-out of GM's software and services organization has been bumpy, as evidenced by problems with the new Chevrolet Blazer EV and by this month's elimination of more than 1,000 jobs globally.

"As we help build GM's bold vision for the future, we must make bold choices that enable us to move faster, pivot when needed, and prioritize investing in what will have the greatest impact," Baris Cetinok and Dave Richardson, two Apple alumni appointed in June to lead the organization, wrote in an Aug. 19 memo viewed by Automotive News. "The resulting changes reflected these exact things: shifting resources to our highest-priority work, simplifying team structures including flattening hierarchies, and avoiding duplication."

GM has heightened its focus on identifying glitches earlier in the software development process through a quality division created after it halted sales of the Blazer EV in December to fix malfunctioning screens and charging errors.

"Software strategy and software quality, it's not ending. We're going to surpass probably 120 million lines of code. And the vehicle is a very complex creature," said Abdul Bazzi, GM's vice president of software quality. "We're making tremendous progress. We have the right talent, we have the right team. We're going to get there."

Bazzi's team operates out of GM's Global Technical Center in Warren, Mich. It's focused on creating centralized testing and validation processes to ensure software works correctly when launched, he said. GM has infotainment-focused testing labs in Michigan, California, Canada and Israel. A second lab in Michigan focuses on software integration.

In the past, GM had employees responsible for software quality within certain focus areas, but no one had end-to-end oversight, Bazzi told Automotive News during a visit to a software quality lab. Now, all decisions about when software applications are ready to be integrated into a vehicle and launched into production run through his team.

The software quality lab in Warren was created last fall with 70 test benches that mimic the infotainment systems found in GM vehicles. It since has more than quadrupled that to 300, said Amy Talerico, GM's director of infrastructure, labs and fleet.

A staff of seven engineers maintains the labs, though operations are heavily automated with an artificial intelligence tool the team developed to run simulations around the clock.

Bazzi's team reviews all software applications for any issues before deciding whether they are ready for further testing or to launch. Infotainment systems are among the most involved because they interact with so many other parts of a vehicle, he said. If a glitch slips through, the AI tool can be trained to look for it in the future. Bazzi said he also monitors warranty issues reported at dealerships weekly to identify burgeoning problems, adding that the trend is moving in a "really good direction."

"One of the things that we're doing is ensuring that we go back and we have the right testing in place. Why didn't we catch it?" he said. "We continue to build our automation. We continue to make our automation much more intelligent, leveraging AI as much as possible."

Bazzi said his team is working to gain the ability to evaluate software virtually, before it gets to a physical test bench, so developers can identify and fix problems earlier.

"What they're doing at the lab is extremely important," said Sam Abuelsamid, principal research analyst at Guidehouse Insights, who has toured the lab.

The combination of manual and automated testing allows GM to run thousands or even millions of simulations that emulate someone tapping on a screen or pressing buttons, he said.

Decades ago, vehicle software was simpler and not expected to change after the car was built. Today, automakers want the ability to send updates over the air after a sale. Software also is more complex because more components interact now. For instance, an infotainment screen once might have interfaced only with the radio but now needs to pull in data on traffic conditions, route maps and an electric vehicle battery's charge.

That all requires different skill sets, tools and processes, Abuelsamid said. GM is learning what is needed for a next-generation software platform and realigning its workforce to match.

If automakers don't meet customers' software expectations, "then consumers are going to go to the manufacturer that can deliver the best experience for them in the vehicle," he said. "Even if they never generate a penny of subscription services, just having that functionality, that expectation of functionality on modern vehicles, is so different than before."