From Skeptic to Standard: How Amazon Engineers Learned to Work With AI

Anni Chen, a technical lead at Amazon, didn't trust AI to write code. The idea clashed with the careful craftsmanship she’d built her career on. Yet today, she uses AI coding tools daily. Her shift from doubt to dependence reflects a quiet revolution reshaping software engineering.

The practice, sometimes called 'vibe coding,' involves describing a task in plain language and letting an AI—like Amazon's own Q Developer or GitHub's Copilot—draft the actual code. The engineer then reviews, guides, and refines. Chen’s initial fears were common: would this erode essential skills or hide dangerous bugs? Her mind changed with the results. Repetitive work—boilerplate code, unit tests, updating old modules—that once took hours now finishes in minutes.

This isn't just one engineer’s story. Amazon CEO Andy Jassy has noted a significant portion of new company code now comes from AI. Google reports over a quarter of its new code is AI-generated. The engineer's role is evolving from writing every line to directing the process.

Significant questions remain. Critics point to risks in security and performance, especially for critical systems, where AI can produce code that looks right but fails in unexpected ways. Legal battles over what material these AI models were trained on continue, leaving copyright questions unresolved.

For now, the industry is pushing ahead. At Amazon, the tools are framed not as replacements but as amplifiers, saving thousands of engineering hours. The demand for engineers who can design systems and ensure reliability remains strong, even as the entry-level path into the profession may narrow. Chen’s journey suggests the new imperative: use the tool, but never stop checking its work. The human role is becoming less about writing syntax and more about ensuring the result is sound.