Andrew Ng on AI-native teams
- Andrew Ng said on April 24 that “AI-native” software teams work differently, with engineers taking on product, design and marketing duties as coding speeds up. - In his DeepLearning.AI letter, Ng said teams of roughly two to 10 people can move fastest, while coding is no longer the main bottleneck. - The argument extends Ng’s recent writing on agentic software workflows and hiring pressure in AI engineering (deeplearning.ai)
Andrew Ng said AI is changing software teams by shrinking them and widening what engineers do beyond writing code. (deeplearning.ai) In an April 24 letter published by DeepLearning.AI, Ng wrote that “AI-native software engineering teams operate very differently than traditional teams.” He said coding agents let teams build products much faster. (deeplearning.ai) Ng said some engineers now act partly as product managers, designers and sometimes marketers, because software can be produced faster than adjacent decisions can be made. He wrote that the bottleneck is shifting away from typing code and toward product definition, design choices and other non-coding work. (deeplearning.ai) He also argued that very small teams can move fastest in this setup. His example was teams of about two to 10 people, especially when they share an office and can communicate face to face. (deeplearning.ai) That pushes hiring toward broader operators rather than narrow specialists. Ng wrote that AI-native development “needs generalists,” because the work now spans coding, product judgment, design trade-offs and rapid iteration. (deeplearning.ai) The argument fits with Ng’s recent focus on “agentic AI,” his term for software that completes multi-step tasks through planning, tool use, reflection and collaboration. DeepLearning.AI’s current course materials describe those workflows as a new way of building software rather than just generating one-off answers. (deeplearning.ai 1) (deeplearning.ai 2) Ng has been returning to the labor side of that shift in recent weeks. On March 27, he wrote that many people at different seniority levels were reporting job insecurity as AI tools changed expectations around software work. (deeplearning.ai) His April 24 note does not claim specialist roles disappear. It argues that when coding gets cheaper and faster, the scarce resource becomes deciding what to build, how to shape it and how quickly a small team can align around those calls. (deeplearning.ai) The throughline in Ng’s case is simple: if AI compresses the time needed to produce code, team structure starts to matter as much as model quality. (deeplearning.ai)
