74% of Employers Now Want AI Skills. Only 1 in 6 Will Train You.
More employers now expect AI skills than are willing to teach them. ZipRecruiter’s 2026 AI Employer Report found that 74% of employers call AI skills a strong advantage or an outright requirement for at least some roles, and PwC quantifies the payoff: jobs asking for AI skills are growing 69% versus 9% for the market overall, and the average wage premium for those skills has risen to 62%. The bar moved this year, and the training did not.
Only about one in six companies runs employer-provided AI training, and that share has barely moved. Workers noticed and stopped waiting. The share learning AI on their own rose from 22% to 30% in a year, and 47% of job seekers worked on their AI skills in the past six months. If you are looking for work this fall, that gap is the whole story: the skill is now expected, the employer will probably not supply it, and the people who build it themselves are the ones who clear the bar.
The bar went up, and it went up this month
This is not a slow drift. iCIMS’s September 2026 report found openings still outpacing hires, with openings up 13% year over year against hires up just 2%, roughly 30 applicants per opening, and about 40 days to fill a role. Employers are adding AI skill requirements into that market, in the same month, while their own training programs stay flat.
The numbers behind the expectation are large enough to be worth stating plainly. A 69% growth rate for AI-skill postings against 9% for everything else means the requirement is spreading roughly eight times faster than the overall job market. And the 62% wage premium is not a rounding error. It is the market putting real money on a skill most candidates have not yet built.
This went out to subscribers yesterday morning.
The daily issue lands at 6am with one story like this and one thing to do about it. Free.
Why a little AI skill goes a long way
Most candidates have surface exposure and nothing more. iCIMS found that 61% are familiar with general tools like ChatGPT, Copilot, or Gemini, but only 18% claim prompt engineering and just 17% can build a machine-learning model. That thin middle is your opening.
You do not need to become an engineer. The distance between “I have used ChatGPT” and “I can show you something I built with it” is a weekend of focused work, and it is the distance that separates you from most of the pile. A single demonstrable skill, applied to one real problem, puts you ahead of candidates with more years and less proof.
It is also becoming a question job seekers ask employers, not the other way around. In a survey of 1,000 U.S. job seekers, 42% said an employer offering AI training is more attractive than one that does not, and about 14% would take a lower salary to get it. Training has moved from a perk into a screening criterion. That changes the interview in your favor, if you use it.
This is not a cure-all
Learning AI will not fix every door, and the entry level feels it most. ZipRecruiter found that 38% of employers have shifted basic data-processing work off entry-level staff and onto AI, and 31% raised experience requirements for entry-level jobs as a result. PwC found that AI-exposed entry-level roles are seven times more likely to demand senior-level judgment, and that those roles grew 35% since 2019 while other entry-level roles fell 10%.
That is why fluency alone is not the answer. Employers are weighting human skills higher too. According to the same reports, critical thinking (65%), judgment and decision-making (59%), and creativity (58%) are all more important than a year ago, alongside technical capabilities like workflow automation and data analysis at 60%. Pair the tool with the thinking. That combination, not the tool by itself, is what the market now rewards.
What to build, and what to show
Build one thing you can show, not a line on your résumé. Pick a real task, automate the boring part, and write down the hours it saved. “I built a workflow that cut our weekly reporting from three hours to twenty minutes” beats “AI skills” every time, because it is specific, checkable, and describes an outcome rather than an interest.
Plain honesty about what the tool did and what you verified is what makes it credible in the interview. Say which parts you automated, say which parts still need a human, and say what you would do differently. That is what a hiring manager hears as judgment, and judgment is the thing the entry level is now being asked to demonstrate.
Learn it cheaply, and use it as a filter
You do not need a certificate to clear a bar this low. The free tiers of the major tools, one focused project at work, and a little deliberate practice are enough to reach the level most postings actually mean. A credential proves you attended something. A working example proves you can do something.
Then make training a question you ask them. “How does the team use AI today, and what training do you offer?” It signals that you are current, and it screens the employer on the one benefit 42% of job seekers now weigh. The answer is informative either way. A company with a real answer is telling you what it invests in. A company with a shrug is telling you something too.
Workers have already said what they want here. In the Express Employment Professionals and Harris Poll research, 83% of U.S. job seekers said companies need to formally train people on AI, and 86% of hiring managers agreed. Only 36% of companies hand employees a list of approved tools. So expect a gap between what employers say about AI and what they actually provide, and treat that gap as information rather than a reason to walk away.
Where this fits in your search
The AI-skills gap sits alongside the other shifts reshaping this fall’s market, including the 2026 job-market trends that actually moved. It also interacts with the application problem, since a specific, demonstrable skill is one of the few things that survives the volume that swallows most cold applications.
People are searching “What AI skills are employers looking for?” and “Is it too late to learn AI in 2026?” The answer to the second one is no, and it is not close to late. The bar is real, it pays, and most of the people standing next to you have not cleared it yet.
Joshua Waldman wrote Job Searching with Social Media For Dummies after being laid off twice in six months (2008–2009). He’s changed careers five times and learned firsthand that technology is the job seeker’s biggest advantage. At CareerEnlightenment he shares what actually works.
