Microsoft has begun asking its engineers to use both Anthropic’s Claude Code and its own GitHub Copilot, opening a rare internal comparison between rival AI coding assistants.
The experiment reaches into core product teams working on Windows, Microsoft 365 and Teams. Engineers now test multiple tools, measure output and share feedback on speed, accuracy and usefulness.
The move reflects a practical question many professionals face at work: stick with familiar tools, or trial alternatives that might deliver better results. Microsoft chose the latter, even when that means putting an external product alongside one of its own.
The company has pushed the experiment further by encouraging non-developer staff to use AI tools for basic coding and prototyping. Designers, managers and analysts can now turn ideas into working demos without waiting for engineering time.
That shift carries clear implications:
- AI lowers the barrier to entry for technical work inside large organisations.
- Software creation spreads beyond traditional developer roles.
- Productivity gains depend less on who writes code and more on who frames the right problem.
GitHub Copilot remains Microsoft’s flagship coding assistant, deeply integrated into its developer ecosystem. Testing Claude Code alongside it signals a willingness to prioritise outcomes over exclusivity.
The approach resembles how businesses evaluate suppliers during a market shift. When conditions change, loyalty gives way to performance reviews. Microsoft appears to be applying that logic internally.
The broader question sits beneath the experiment. If AI allows more employees to write functional code, how does that reshape the role of professional developers? Teams may shift focus from writing routine code to reviewing, securing and refining AI-generated output.
The trial also carries risk. Expanding coding access increases the need for governance, security checks and quality control. A single flawed script can scale quickly when generated by AI.
Microsoft’s decision suggests confidence that the benefits outweigh the costs. The company is not betting on a single tool or vendor. It is betting that AI-assisted development, used widely and pragmatically, will change how software gets built inside one of the world’s largest technology firms.
If this model works at Microsoft’s scale, how long before it becomes standard practice elsewhere?
Author: Pishon Yip