AI Skills Go Mainstream, but the Training Gap Is Real

Two big pieces of jobs research landed in the last fortnight, and read together they tell one of the most important stories in work right now. AI skills have gone mainstream, far beyond the tech team. But the training hasn't caught up, and the people paying the price are the ones at the start of their careers. That's a problem, and it's also one of the biggest opportunities I've seen for anyone willing to get ahead of it.

AI skills are now an everyday job requirement

ZipRecruiter's 2026 AI Worker Report, published on 26 October, has the headline number: job postings that call for using AI grew 159% year over year, triple the 53% growth in jobs that develop and deploy AI. Non-tech roles that mention AI now make up 4.3% of all postings on ZipRecruiter, up from 1.9% a year ago. And AI infrastructure roles, such as data centre technicians, are up 169%.

The weekly AI Pulse tracker points the same way from a different dataset. Across 142,049 active postings, 17.6% now require AI, up from 15.4% in November 2025. Engineering leads at 45.5%, but product (38.6%), data (38.2%), marketing (21.8%), legal (17.8%) and finance (15.9%) are all well into double digits.

I love this, because it's the clearest sign yet that AI has stopped being a specialist niche and become a basic professional skill, like spreadsheets in the 1990s. You don't need to be an ML engineer to benefit. You need to be the marketer, accountant, recruiter or operations lead who knows how to get real work done with these tools.

Who's ramping up AI hiring

It's not just the AI labs. Level's AI hiring signals tracker, updated 1 November, ranks companies by the rise in jobs that work on or build AI. The biggest risers over the last 30 days are not the usual Silicon Valley names: Thales (241 jobs, up from 58), Airbus (128, up from 14), GSK (117, up from 20), Cisco, Intel, ElevenLabs, Wells Fargo, NVIDIA, Harvey and Graphcore. Defence, aerospace, pharma and banking are all moving.

The AI Career Atlas hiring index for November adds colour on the type of roles being created. It counted 9,417 open roles at 80 AI companies on 1 November, up 213 on October among companies measured in both months, and its list of new titles is a fascinating read: Databricks has added an "AI automation engineer", Crusoe is hiring electrical design engineers for modular data centres, and Nebius is hiring data centre technicians. The AI economy is creating jobs in sales, finance, operations and the trades, not just in research.

The training gap is real

Here's where it gets uncomfortable. ZipRecruiter surveyed more than 1,500 US workers in September and found 82.5% use AI in some form at work, and 52.6% say it's made work more enjoyable (only 5.5% say less). That's great news. But only 28.9% say they get enough training, and 37.2% are learning entirely on their own because their employer provides none.

The split by seniority is stark. 83.5% of senior leaders report having AI training at their company, compared with 46.4% of individual contributors and 45.3% of entry-level workers. Unsurprisingly, 76.5% of senior leaders say AI has improved their work satisfaction, against 39.8% of entry-level staff. And 54.7% of entry-level workers are teaching themselves.

Meanwhile expectations are rising. 49.9% of workers say their employer has raised productivity expectations because of AI, and 44.6% say AI is making their technical skills go out of date faster. ZipRecruiter labour economist Nicole Bachaud put it neatly: "As expectations rise, the workers without access to formalized training are the ones most at risk of falling behind."

Why early-career workers feel it most

The ADP Research and Stanford Digital Economy Lab Canaries Dashboard, published on 21 October, shows the same pressure in payroll data. Employment in occupations highly exposed to AI shrank 0.5% in September from a year earlier, while the least-exposed occupations grew 1.0%. For workers aged 22 to 25 in high-exposure jobs, employment fell 3.7%, the 36th straight month of year-over-year contraction.

ZipRecruiter offers a plausible reason. The tasks workers hand to AI most often are research (52.1%), summarising (41.6%) and drafting text (40.9%). Those are exactly the tasks that used to teach new hires how a business works. When the training ground gets automated and the formal training doesn't arrive, the first rung on the ladder gets harder to reach.

I want to be careful here, because this is not a story of mass job losses. S&P Global's AI and Labor Landscape 2026 report does show AI's net employment effect turning modestly negative, at -5 points over the past year and a forecast -2 points for the coming year. But it also found that companies prioritise process efficiency (64%) and productivity (59%) far more than head count reduction (24%), and only 22% of AI projects aim for a fully autonomous end state. The overwhelming picture is people working with AI, which means the people who use it well win.

What I'd do if I were job hunting right now

The good news is that the fix is mostly in your own hands, and it's cheap. Here's my practical playbook.

Stop waiting for your employer to train you. ZipRecruiter career expert Sam DeMase said it plainly: "Job seekers can't wait for their next employer to provide them with AI training." Pick the tools your target industry uses (ChatGPT, Claude, Gemini, Microsoft Copilot and Grok Bot are all good places to start) and use them on real tasks every day for a month.

Build workflows, not just prompts. The report found 36.1% of workers configure existing AI tools or plugins, while 31.6% haven't customised anything at all. Moving from "I ask it questions" to "I built a workflow that turns our weekly sales data into a board summary in ten minutes" is the difference between a line on a CV and a story that gets you hired.

Show, don't tell. Put two or three before-and-after examples in your portfolio or LinkedIn: the task, the time it used to take, how you use AI now, and how you check the output. Hiring managers want evidence of judgement, not a list of tools.

Double down on the human skills. 40.4% of workers say they're focusing on human skills to future-proof their careers, and they're right. Critical thinking, checking AI output, client relationships and decision-making are what AI makes more valuable, not less.

Look where the growth is. AI infrastructure roles are up 169% on ZipRecruiter, and the AI Career Atlas new-titles list is full of operations, finance, sales and data centre roles at AI companies. You don't need a PhD to work in AI.

What employers should take from this

If you hire or manage people, this research is a gift. 40.2% of workers say they could master AI skills with proper training, and 25.1% say they're planning to leave their job because of how their employer uses or views AI. That's a retention risk and a productivity opportunity in the same sentence. Hands-on, role-specific training beats a generic overview for the C-suite every time, and ZipRecruiter's own Chief People Officer, Amy Garefis, makes the point that peer-to-peer sharing of real use cases is some of the most valuable learning there is. And please, keep hiring juniors. Give them AI tools and a mentor, and they'll get up to speed faster than any generation before them.

My take

I'm optimistic. Every one of these datasets shows AI expanding the number of jobs that touch the technology, and companies in defence, aerospace, pharma and banking racing to hire for it. The training gap is real, but it's also an open door: the people who teach themselves now, and can prove it, will be the ones employers fight over in 2027.

If you're ready to put those skills to work, browse live AI and automation roles on our jobs board, from agent engineering to AI operations, and find the one that fits where you want to go next.