Careers that are growing BECAUSE of AI

A former coworker messaged me in April. She's a technical writer, eight years in. She'd just accepted a job with a 42% pay bump and a new title: "AI documentation lead." Her actual work sounds a lot like her old work — writing docs, owning style, pushing back on bad API design — except now she also writes the prompts, examples, and eval cases that ship with the company's SDK. The job didn't exist two years ago. She didn't invent it. Three different companies tried to hire her for some version of it.

That's the pattern I want to talk about. Not the "AI is destroying jobs" story or the "AI is creating jobs" story — both are too tidy. The real thing is that specific existing roles are getting a second wind because the new tools need someone to think carefully about them. Here's where I see it happening.

Integration engineers (the people who plug AI into actual products)

Every mid-sized company wants to "add AI" to their product. Almost none of them have anyone who has shipped an AI feature that didn't embarrass them. So they pay a premium for people who have. This role is mostly software engineering with extra requirements: you understand latency budgets, token costs, eval frameworks, and why a demo that works on Tuesday breaks on Thursday.

Salaries I've seen for this in Berlin and Amsterdam: €90k-€140k for senior ICs, and the ceiling keeps rising. A backend engineer with two real AI features under her belt is in a much better negotiating position than one without. If you're already an engineer, this is the most direct upgrade path.

Evaluators and red-teamers

Someone has to answer: "did the model get better this week?" Turns out that's a full-time job. Writing evals, running them, interpreting the results, figuring out which regressions are real and which are noise — this is a job that barely existed in 2022 and now has its own conference tracks.

The funny thing: the best eval people aren't usually ML researchers. They're domain experts who learned enough about LLMs to check the work. An accountant who builds the accounting evals. A lawyer who builds the legal evals. A nurse who builds the triage evals. That's a real move and the pay is competitive with the domain job it came from — often better.

Forward-deployed engineers and solutions architects

Anthropic, OpenAI, and a growing list of AI vendors hire people who basically embed at a customer for weeks or months and build the thing. It's half engineer, half consultant, half teacher. You need to ship code, translate between technical and non-technical stakeholders, and have the stomach for the first demo blowing up spectacularly.

These roles existed before, but the volume has exploded. A friend at a smaller AI startup told me their FDE hiring bar is "senior engineer who can talk to humans without flinching," and they can't find enough of them.

Content and product ops for AI-heavy companies

Less glamorous, more common. Someone has to manage the prompts in production like code — versioned, tested, rolled out carefully. Someone has to own the knowledge base the retrieval system searches. Someone has to write the internal docs that say "here's how we actually use these tools and here's what's banned." This is often labeled "AI ops" or "LLM ops" and it sits between product, engineering, and content.

If you're currently in content ops, knowledge management, or technical program management, this is a lateral move with a vertical salary.

A concrete example of the compound effect

A product manager I know, 31, ex-finance background, joined a SaaS company four years ago as a PM doing nothing AI-related. In early 2024 she volunteered to own their first AI feature — a summarization thing nobody else wanted because it was risky. She spent a year learning, shipping, screwing up, re-shipping. Now she's Director of AI Product, owns a team of seven, and her last offer from a competitor was north of €180k plus equity.

She didn't become an ML engineer. She became the PM who actually understands what the tools can and can't do, which is rarer than you'd think and priced accordingly.

The meta-pattern

The roles growing because of AI aren't mostly "prompt engineers" or "AI researchers." They're existing roles — writer, PM, engineer, analyst, domain expert — with an AI layer bolted on. The people winning are the ones who were already good at the underlying craft and then spent six to twelve months getting genuinely fluent with the tools. Not tool-curious. Fluent.

The uncomfortable truth for mid-career professionals: this probably means you, if you want it. The new roles are mostly being filled by people who were already in the field and decided to go deep on AI. Not by 22-year-olds parachuting in.

Takeaway

The growing-because-of-AI careers aren't exotic. They're your current job plus a focused six-to-twelve-month investment in genuinely understanding the tools. Pick the role adjacent to yours that seems most interesting, find one real project to ship with the tools, and you're already ahead of 90% of applicants for the job that didn't exist two years ago.