Andrew Ng's AI skills list draws fire for ignoring business and governance needs

Ng listed the abilities in a post, emphasizing that all developers will need AI engineering skills. The four areas are building and deploying AI applications, understanding software engineering fundamentals, using coding agents, and shaping the build. For building and deploying, Ng highlighted knowledge of LLMs, context engineering, RAG, agentic workflows, and statistical techniques to govern AI systems. Software engineering fundamentals include architecture, testing, and security, which he said lead to better outcomes than inexperienced developers "vibe coding" without understanding tradeoffs. Using coding agents requires a mental model of how agents work, their limitations, and how to steer them. "Shaping the build" means engineers need product sense, business context, and customer goals rather than just implementing pixel-perfect designs.

However, the framework drew pushback. Andy Thurai, founder and AI advisor at The Field CTO, called it "a dangerously narrow framework for the enterprise" with a "massive blind spot of builder bias." He said Ng's taxonomy is focused on "Day 1" innovation—writing code and getting the model to work—but in enterprises, the hardest part is "orchestrating, observing, and paying for it." Deepika Sidana, senior manager of software engineering at American Express, agreed that the four skills are important but not enough on their own. Engineers also need to understand the business problem, customer workflow, risk, compliance, and the consequences of failure.

Sidana said strong orchestration skills are critical, encompassing coordinating models, tools, data, evaluations, observability, human approvals, and fallback paths. Thurai added that key skills include multi-agent orchestration, AI FinOps, agentic observability, agentic security, and socio-technical systems integration. "Innovation gets you to the starting line, but observability, security, resilience, and governance get you to production," he said. Naman Ahuja, a software engineer at Meta, said the most valuable AI-era engineering skills increasingly sit outside traditional coding.