AI Jobs: What’s Actually Sticking?

Three years after ChatGPT’s public debut in November 2022, the promised artificial intelligence job revolution has arrived. It’s just not what anyone anticipated.

See Also: Live Webinar | AI-Powered Defense Against AI-Driven Threats

Research from Indeed’s Hiring Lab, the research arm of the job posting platform, shows that demand for AI fluency – the ability to use and manage AI tools – jumped nearly sevenfold in two years.

McKinsey Global Institute, the research arm of McKinsey & Company, found that about eight million people in the United States work in occupations where job postings already call for at least one AI-related skill.

But within those numbers are unexpected outcomes. Organizations initially reacted to the chatbot’s debut by rushing to hire prompt engineers – specialists tasked with crafting text instructions that coax optimal responses from AI systems. Job postings proliferated offering salaries ranging from $175,000 to more than $300,000. AI trainers were in hot demand. These roles seemed poised to become cornerstones of the AI economy.

The picture today looks starkly different. Prompt engineering survives on freelance platforms such as Upwork but it has largely vanished as a full-time corporate role. AI governance specialists – professionals who establish policies for safe AI use – are the ones organizations actually want.

Fiona Mark, principal analyst at Forrester, watched this evolution firsthand. Certain roles initially seemed poised for explosive growth have not played out that way, she told Information Security Media Group. Prompt engineering, or context engineering, is a core skill for engineers working with AI, but is no longer considered a standalone skill. It has merged into AI engineer, which covers a wide range of working with large language models and AI platforms.

Andrew Rabinovich, chief technology officer and head of AI and machine learning at Upwork, said that as foundation models became more capable out of the box, demand for bespoke model development and low-level systems engineering diminished.

The pattern extends to programming languages. Ruby, Rust and Scala experienced short-lived surges in hiring on Upwork in 2023 during the early wave of enterprise LLM adoption, before declining sharply, said Rabinovich. Programming languages that briefly surged based on expectations of custom model-building or infrastructure control did not ultimately become core languages for interacting with the modern AI stack, he said.

Python has become the universal interface language of AI, powering data pipelines, model integration, agent frameworks, evaluation workflows and almost every major LLM library, allowing it to continue to be essential, even as models and tools evolve, Rabinovich said.

The most durable roles center on governance and integration, positions that address how organizations effectively and safely use these tools and platforms. AI governance specialists will continue to see growth as AI becomes more embedded in workflows and organizations seek to manage and mitigate risk, said Mark.

Rabinovich identified entirely new categories of AI work that have emerged and are growing quickly, including AI agent development and a broad set of multimodal AI content generation skills, such as image, video and synthetic voice creation. Most emergent AI roles are focused on the integration of LLMs into complex workflows such as agent developers and agent orchestrators, he said. “We expect these will be among the most future-proofed,” he said.

The prevailing differentiator in the AI job market lies in applying domain knowledge and uniquely human skills like creativity, decision-making and judgment to AI implementations and outputs, rather than performing required tasks or mastering tools, Rabinovich said.

There is continued demand for AI domain specific skills, such as AI engineers who build and maintain platforms, AI solutions engineers who design solutions and specialists in agentic workflow design. Data governance specialists, cloud engineers, architects and AI security engineers are also in demand.

Rabinovich sees a stable class of AI operators emerging, ranging from agent orchestrators to professionals handling data labeling and model fine tuning. As AI tools permeate the entire development stack, talent across the spectrum will be essential, he said.

Mark said it’s still an open question whether AI specialists will become widespread or whether most AI skills will become embedded into other functions. “It’s very early to tell,” she said. It’s likely that some specialist roles remain, such as agentic and AI solutions engineers, while certain other skills around model training or integrations become core skills across various domains.