GM redesigned its engineering workflows around AI agents — and tripled its merged pull requests
Source: VentureBeat AI
Summary
- GM's autonomous driving division has seen significant gains after redesigning its engineering workflows around AI agents.
- These agents help with tasks such as analyzing vehicle data, triaging problems, and running experiments.
- As a result, GM has seen faster releases and fewer defects escaping into later stages of development.
- GM's engineers spent most of their time outside the code editor before using AI agents, with tasks like debugging and refactoring taking up a large portion of their time.
- A recent Microsoft study found that developers spent an average of 96 minutes writing code on good workdays, which is only a small part of their overall work.
- GM has given its AI agents access to internal tools and data by connecting them to customized Model Context Protocol (MCP) servers.
- This allows the agents to analyze data, conduct initial triages, and create issues for engineers to investigate.
- The output of the agents must be human-readable, and their permissions are based on those of the engineer using them.
Why It Matters
- GM's success with AI agents shows that companies can achieve huge gains by redesigning their engineering workflows around these tools.
- This means that other companies can learn from GM's example and adopt similar strategies to improve their own development processes.
- Everyday people will benefit from faster and more efficient software development, which can lead to better products and services.
- GM's use of AI agents also highlights the importance of accountability in AI development.
- The engineers using the agents are still responsible for the output, which ensures that the agents are used correctly and safely.
- This approach can help to build trust in AI systems and prevent potential problems.
- GM's experience with AI agents also shows that these tools can be a key driver of innovation in various industries.
- By automating tasks and freeing up developers to focus on more complex and creative work, AI agents can help companies stay ahead of the curve and drive growth.
GenAI EXPLAINED
AI agents are computer programs that can perform tasks independently, often by interacting with other systems and tools. GM's AI agents are designed to work with internal tools and data, allowing them to analyze vehicle data, conduct triages, and create issues for engineers to investigate.
Model Context Protocol (MCP) servers are customized systems that connect AI agents to internal tools and data. This allows the agents to access the information they need to perform their tasks, and to call underlying tools used by other systems.
Permissions are a way of controlling what tasks an AI agent can perform, and what data it can access. GM's AI agents have permissions based on those of the engineer using them, which ensures that the agents are used correctly and safely.
MORE FROM THIS EDITION