GitHub Copilot Upgrades Enterprise Features
GitHub has rolled out new features for its Copilot Business and Enterprise tiers. Administrators can now select the specific AI model that powers the asynchronous coding agent, offering more control over code generation. Additionally, Copilot support for the Zed code editor is now generally available for all paid subscribers, expanding workflow options for development teams.
- The new model selection feature provides administrators with control over which AI models their teams can use, balancing factors like speed, reasoning depth, and cost. Options can be limited for security or governance reasons and may include models like GPT-4.1, GPT-5-mini, and various versions of Claude, each with different performance characteristics. - GitHub Copilot Enterprise is priced at $39 per user per month and requires a GitHub Enterprise Cloud subscription. This tier includes features from the Business plan ($19/user/month), such as IP indemnity and centralized policy management, but adds capabilities like chat personalized for a company's private codebase. - The integration with Zed, a high-performance, Rust-based code editor, is significant as Zed is built for speed and real-time collaboration, positioning itself as an alternative to Electron-based editors like VS Code. Zed offers native AI features and allows users to connect to various large language models, including Copilot. - While AI coding assistant adoption is high, with some reports indicating 85% of developers use them regularly, concerns about the quality and security of AI-generated code persist. Studies show that while developers feel more productive, less than half of AI-generated code is accepted without modification, highlighting the need for human oversight. - The "auto" model selection setting, now in public preview, optimizes for model availability and routes requests to a range of models, including those from OpenAI and Anthropic. For paid subscribers, using the "auto" feature provides a 10% discount on the "premium request" multiplier, which is how usage of more powerful models is billed. - The broader trend for AI in software development is shifting focus from just writing code faster to improving the entire development lifecycle, including tasks like documentation and requirements analysis. However, without strong underlying workflows, AI can amplify existing bottlenecks in areas like code review and testing. - Zed's native support for Copilot was initially enabled by leveraging an undocumented, minified Language Server Protocol (LSP) server from the open-source Neovim plugin. This highlights the community-driven efforts to integrate popular tools before official APIs are available. - The adoption of AI coding tools is projected to continue its rapid growth, with Gartner predicting that 90% of enterprise software engineers will use AI code assistants by 2028, a significant increase from less than 14% in early 2024. This growth underscores the increasing importance for engineering managers to develop strategies for effective AI tool implementation.
