How GM Tripled Pull Requests with AI Agents: A Game-Changer for Engineering Workflows

GM's AI-Powered Engineering Revolution: What Happened

General Motors is reshaping how software engineers work, and the results are striking. According to insights shared at VB Transform 2026, engineers at GM's autonomous driving division now spend only 15% of their time actually writing code. The remaining 85%? That's now being accelerated by AI agents handling tasks like analyzing vehicle data, triaging problems, running experiments, and managing other critical workflows.

The outcome speaks for itself: GM has tripled its merged pull requests, a key metric showing development velocity and code quality. This isn't just a productivity bump—it's a fundamental reimagining of how modern engineering teams operate.

Why This Matters for the AI Landscape

GM's success story provides crucial real-world validation that AI agents can transform enterprise engineering workflows at scale. This is particularly significant because:

• Workflow automation is reaching critical mass: AI agents are moving beyond chat interfaces into deeply integrated development processes
• Enterprise adoption is accelerating: When a company as established and complex as GM restructures around AI agents, it signals broader industry readiness
• Metrics-driven proof: Tripling pull requests isn't anecdotal—it's measurable evidence of AI's impact on engineering velocity

What Changed in GM's Workflow?

The autonomous driving division didn't just add AI tools on top of existing processes. They fundamentally redesigned engineering workflows around AI agents. This means:

• AI agents now handle data analysis that previously required manual engineering effort
• Problem triaging—identifying and categorizing issues—is now automated
• Experiment running and iteration cycles are faster with AI assistance
• Engineers focus their time on high-level problem-solving and creative work

This shift reflects a broader trend in AI tooling: the best implementations don't simply replace tasks—they restructure entire processes around what AI does well.

Implications for AI Tool Users

If you're evaluating AI tools for your engineering team, GM's experience offers several lessons:

• Integration matters more than features: Standalone AI tools won't deliver GM's results. Deep integration into your actual workflow is essential
• Think about workflow redesign: Don't retrofit AI into old processes. Consider how AI agents could fundamentally change how your team works
• Measure what matters: Track metrics like merged pull requests, cycle time, and engineer productivity to validate ROI
• Focus on high-impact tasks: GM's AI agents handle the 85% of non-coding work. Identify your equivalent efficiency bottlenecks

The Bigger Picture

GM's story demonstrates that AI agents are moving from experimental tools to production-grade systems capable of handling complex, mission-critical work in highly regulated industries. The autonomous driving division operates under intense pressure to deliver reliable, safe code—if AI agents can accelerate that, they can likely accelerate engineering anywhere.

This also signals a shift in how companies will compete on engineering talent. Organizations that successfully integrate AI agents into workflows will get more output from their existing teams, changing the economics of software development.

The Takeaway

GM didn't just adopt AI tools—they redesigned their engineering culture around them. For organizations considering AI agent integration, the lesson is clear: the biggest gains come not from incremental tool adoption, but from rethinking workflows from the ground up. As more companies follow GM's lead, we're likely to see AI agents become as fundamental to modern engineering as version control itself.

Source: VentureBeat