Elton Chan is the Co-Founder of Second Talent, a solution that connects global tech leaders with AI-native engineering talent across Asia.
getty
Three years ago, the job title "AI engineer" barely existed. Today, it has split into 10 different roles. The World Economic Forum's Future of Jobs Report 2025 ranks AI and machine learning specialists among the three fastest-growing roles in the world, projecting roughly 85% growth by 2030.
The Stack Overflow Developer Survey 2025 also shows 84% of professional developers now use AI tools at work, but fewer than 5% of developers identify as AI engineers, data scientists or ML specialists, even as many use AI tools daily. The gap between those two numbers is the problem.
Most companies are still trying to fill that gap with one hire. They post a job called "AI engineer." They list eight things the person should do and wait. The role sits open for months, and they wonder why.
The reason is that "AI engineer" has become an outdated job title, and there are several AI engineer roles that require fundamentally different skill sets, workflows and operating models.
This split has happened before. In the 2010s, "web developer" became front end, back end, mobile, DevOps and SRE. Companies that hired specialists shipped faster than companies that kept hiring one generalist, and the market rewarded the split.
The AI version is faster. I first started to see stable AI engineer job descriptions appear in early 2023. By mid-2024, the role had already split into LLM engineer, prompt engineer, RAG engineer and AI agent engineer. In 2026, other entries include AI automation engineer, workflow automation engineer, LLMOps engineer, AI product manager, AI solutions architect and AI safety engineer.
Each one is a real job. The AI automation engineer wires AI into operational workflows. The RAG engineer owns the retrieval layer. The AI agent engineer designs autonomous systems with tools such as LangGraph and MCP. The AI safety engineer red-teams it all. A senior engineer can maybe do a couple of these roles well, but almost none can do them all.
Most hiring teams have not felt the split yet. CEOs read about AI in a board update and the CTO knows the product needs to ship, so HR pulls last year's template, swaps in some keywords and posts the role. The result is a job description that lists eight specializations and a salary that pays for one.
What follows is predictable. Strong candidates do not apply. The job looks impossible, and the applicants who do show up are generalists with light depth on every line. After six months, the new hire is shipping slowly and hitting limits on the harder problems. The role gets rewritten, and the cycle repeats.
The fix is not to hire harder. The fix is to stop asking for one AI engineer and to understand which specific AI engineer roles your job requires.
A more effective approach is to build AI teams in layers based on the stage of the business and the complexity of the product. A Series A company shipping its first AI feature may only need two core hires: an LLM engineer to build the integration and an AI automation engineer to connect AI into operational workflows. Hiring for niche specializations too early often creates overhead before the company has enough production complexity to justify it.
As products mature and customer usage expands, the team structure changes. A Series B company with AI features in production may need a RAG engineer to improve retrieval quality, an AI product manager to manage evaluation-driven road maps and an LLMOps engineer to introduce versioning, monitoring and deployment discipline. Companies building autonomous workflows may also need an AI agent engineer as orchestration and tool use become more central to the product experience.
By the time a company reaches enterprise scale or begins selling into regulated industries, additional specialization becomes necessary. AI safety engineers help reduce reliability and compliance risks, while AI solutions architects align increasingly complex AI systems with customer environments and operational requirements.
This is how mature AI org charts are increasingly taking shape, because AI hiring is no longer a single-role problem. The organizations moving fastest are identifying which AI capabilities matter most at their current stage, hiring for those specific gaps first and expanding specialization only as product and operational complexity increase.
If you are a CEO or board member, stop asking your team about "hiring an AI engineer." Start asking which specializations they need in the next 12 months, in what order and the costs of each. Treat AI as a function with sub-roles, the same way you treat marketing or finance.
"AI engineer" was a useful shorthand in 2023. In 2026, I believe it's a liability. The companies that retire the title and replace it with a real org chart will be better prepared to build the AI systems their business needs.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。