Andrew Ng (응), a Stanford University professor seen as one of the world’s leading AI scholars, posted a write-up on how AI-native software teams operate differently from traditional teams.

In a post he recently shared on social media platform X (Twitter), Ng wrote that the biggest difference is that AI-native engineering teams use coding agents to build products much faster.

As development speeds up, team structure, roles and decision-making also change. Ng said, "Great engineers go beyond writing code and take on product manager, designer and sometimes marketer roles. Small teams working face-to-face in the same space can move at an almost unbelievable speed."

When teams can build quickly, deciding what to build becomes important. Developers also need to spend more time on this. Some engineering teams are adjusting the ratio of engineers to project managers (PMs) from 8 to 1 to 1 to 1 to remove planning bottlenecks.

Ng said there is a better path. He said, "If one PM decides what to build and one engineer builds it, communication between the two becomes a bottleneck. The fastest-moving teams have engineers who can do some product work. When engineers understand users, decide what to build, and can build it themselves, execution speed becomes overwhelmingly faster."

As coding speeds up, bottlenecks can emerge not only in planning but also in design, marketing and compliance processes. Ng said, "Some teams build features so fast that marketing organisations scramble over how to communicate them to users. If a team builds software in a day but the legal team takes a week to review it, the legal review becomes the bottleneck."

Ng said agentic coding is changing not only software engineering workflows but also all surrounding teams. In this environment, generalists carry more weight in small AI teams. He said this is because companies used to bring together specialists in engineering, product management, design, marketing and legal, but if a two-person team must do work requiring five areas of expertise, each person needs more than one specialty.

Ng said, "The tech industry has more engineers than PMs, but both paths are possible," and urged engineers to learn product management capabilities and PMs to learn how to build themselves.

Ng also said it is important for engineering teams in the AI era to work in the same space. Remote teams can do well, but the fastest speed is possible when everyone can communicate immediately in the same room. He said, "I know these role changes are not easy for many people, but I also see hope in that individuals and small teams willing to learn these capabilities can accomplish far more than before. This is a golden era for learning and building."