What jobs will LLMs create, not just replace?

LLMs are creating jobs around building them, checking them, and feeding them — plus more time for the human work that was never going away in the first place.

Every big tech shift works this way: automation kills some tasks and spawns new roles nobody had a job title for a few years earlier. This one is already visible in job postings.

What new job titles are appearing around LLMs?

A cluster of roles has formed around building these models, checking their output, and supplying the data they learn from:

• AI engineers who build and wire LLM features into products
• AI evaluation and safety specialists who stress-test models for bias, security holes, and outright failures
• Prompt and context engineers embedded on product teams, tuning how humans and models talk to each other
• AI trainers who generate the human feedback data (the RLHF work) that models learn from
• Auditors and fact-checkers who review AI output at scale before it reaches customers

What about jobs that don't get a new title?

There's a less obvious category: no new job title, just more of the job you already have. When an LLM handles the repetitive drafting, summarizing, or first-pass research, the person doing that job gets freed up for the parts that actually needed a human — judgment calls, building trust with a client, or hands-on work no model can do yet. A marketer, lawyer, or analyst doesn't become an "AI worker"; they spend less time on the routine middle of the job and more on the ends that were always the point.

Does this cancel out the jobs LLMs replace?

No — and it's important not to oversell it. None of this erases the jobs LLMs are displacing, and there's no guarantee the new roles land in the same place, or for the same people, as the ones that go. It just means the ledger has two columns, not one: some work disappears, some work is created, and the honest picture is the net of both.