The future of AI in software development

Inbal Shani, Chief Product Officer at GitHub and former AWS and Microsoft general manager, discusses how AI is reshaping software development, including Copilot's design, adoption, and measurement. The conversation also covers how product teams should approach AI adoption, how GitHub's research team operates, and her leadership lessons.

01Key learnings

• Engineers using AI tools need to shift from writing code to systems and architecture thinking, which lets junior developers spend more time understanding the product and environment.
• Don't adopt AI for its own sake; start from the customer problem and ask how AI can best solve it, rather than asking what to do with AI.
• AI tools must fade into the background: any extra setup, asking, or waiting adds friction and developers will abandon the tool.
• Measure AI impact with a combination of metrics such as code quality, security, and time to value, with developer happiness as the ultimate outcome, since raw time saved or lines of code are poor proxies.
• Dogfood your own AI features internally across all functions before shipping them to customers, so you learn whether they truly work.
• Give innovation teams a real bridge to production: keep research teams connected to product and engineering so ideas get built rather than becoming academic papers.
• Expect AI-driven testing to become more important as more code is generated, covering unit, load, security, and infrastructure testing.
• When driving change in a new role, first explain the why and bring the team along; moving too fast without context causes resistance.

02Frameworks mentioned

Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.