General Motors: AI Agents Triple Merged Pull Requests in Engineering

General Motors has redesigned how its engineering teams work by building AI agents into their workflows, and has recorded a threefold increase in merged pull requests. The Detroit automaker is now one of the most aggressive examples of AI adoption in software development among large industrial companies outside the tech sector.

The 300 percent increase in merged pull requests means GM's teams are now pushing roughly four times as many finished code changes through review and deployment as before. It's one of the most concrete, measurable indicators of how coding agents are changing the pace of work inside a large engineering organization, not just at software startups.

Scale of GM's Rollout

The most advanced deployment is in the team working on autonomous driving, where, according to Mary Barra, nearly nine out of ten lines of code are now written with the help of AI tools. That is an unusually high share for an automaker, where software is responsible for vehicle safety on the road, not just consumer-facing features.

Alongside code generation itself, GM has also expanded automated testing. The company describes AI-based testing labs that simulate millions of driver and passenger interactions with vehicle interfaces before an update goes into production. This is meant to offset the risk of introducing large volumes of machine-generated code into systems that control cars on the road.

Personnel Context

The overhaul of GM's engineering processes coincides with a deep reorganization of the company's technology team. In recent months the company has lost, among others, its first-ever Chief AI Officer Barak Turovsky, who was hired in March 2025 and departed after just nine months, and software vice president Baris Cetinok. Their replacements bring experience from companies such as Apple, aimed at anchoring GM more firmly in an AI-native approach.

Earlier this year GM laid off hundreds of IT department employees while simultaneously announcing openings for similar roles, but with requirements around working with AI tools. The company is also actively recruiting engineers who specialize in designing AI agents, which suggests the overhaul of its development processes is part of a broader strategy rather than a one-off experiment.

What It Means for the Industry

GM's example shows that automating software development with AI agents is no longer the exclusive domain of tech companies and is spreading into heavy industry, where software is responsible for the safety of physical products. The threefold increase in code review throughput, combined with continued investment in extensive automated testing, suggests the company is trying to balance speed with quality control rather than simply maximizing the number of merged changes.

For Polish automotive companies and suppliers, including the numerous R&D centers of large corporations operating in Poland, this example could become a reference point when planning their own rollouts of coding agents. One question remains open: whether the rise in merged pull requests actually translates into fewer defects and faster feature delivery, or simply shifts the work from writing code to reviewing it.