Job Seekers Are Teaching Themselves AI Faster Than Employers Will Train Them

The clearest jobs data published this week may be the least dramatic. On 10 September, talent acquisition platform iCIMS released its Insights September 2026 Workforce Report, drawing on more than 3.1 million users and 691 million candidate profiles. The finding that stands out: 47 percent of US job seekers said they had actively built AI skills in the past six months, up from 41 percent a year earlier — while employer-provided AI training stayed roughly flat at about one worker in six.

🔍 THE BOTTOM LINE: Workers are not waiting for their employers to upskill them; they are teaching themselves AI faster than companies are building training programmes. That gap is now measurable in hiring, where 42 percent of job seekers say an employer offering AI training beats a comparable one that does not.

The Numbers, And What Moves Them

The self-teaching trend is accelerating. The PR Newswire release reports the share of job seekers teaching themselves AI rose from 22 percent to 30 percent in a single year. Meanwhile, employer-sponsored training “remained stagnant” — a phrase the report uses without irony.

Demand is real. Lightcast posting data cited in the same report shows employers keep adding AI skill requirements across industries, and 45 percent of job seekers said generative AI competencies now appear in descriptions for roles they would consider. But the market is tight in an unusual way: job openings rose while hiring fell for a second consecutive month, and the average applicant pool shrank to about 30 candidates per opening. Trent Cotton, iCIMS’s head of talent insights, called it “a constrained talent market — demand rising, supply flat, and the only lever left is execution.”

The skills people are building are mostly general-purpose. 60 percent of respondents felt prepared to adapt to AI at work, but 61 percent described their proficiency as limited to tools like ChatGPT, Copilot or Gemini — far fewer can do prompt engineering, machine learning or model development. As Intelligent CIO’s commentary put it, employers face two connected problems: accelerate internal AI training, and get better at spotting genuine capability in candidates rather than resume keywords.

The Positive Signal Hiding in a Pessimistic Headline

Read the other way, this is a story about workers choosing to adapt rather than waiting to be displaced. Nearly half of job seekers invested their own time in AI skills within six months. One in seven — 14 percent — said they would accept lower pay in exchange for an employer that trains them. That is not a workforce in retreat from AI; it is a workforce bidding for entry into it, on its own initiative and often at its own expense.

The willingness is not unique to the US. A separate eDreams ODIGEO survey published this month of 9,000 workers across seven markets found 61 percent of US employees would welcome employer-led AI training, with appetite strongest in Portugal at 79 percent. Workers are asking to be trained; most employers have not answered.

What It Means for New Zealand

New Zealand’s mirror image is striking. As we covered last week, KPMG found only 24 percent of New Zealanders have had any AI training while 91 percent of Kiwi workers already use the tools — the same shape as iCIMS’s US data, arrived at from the employer side. The New York Fed’s September survey adds the employer counterpoint: about a third of AI-using service firms retrained existing staff, and researchers called retraining “the primary way firms are adjusting their workforces” — which suggests the flat training numbers are a lag, not a permanent ceiling.

For Kiwi workers the practical read is blunt: if employer training is scarce, the self-taught route is the main road, and it is already crowded. For employers, iCIMS’s data suggests training is now a recruiting tool — cheap relative to the wages war it can replace. The 42 percent preference figure implies a company that adds a structured AI pathway to its job ads competes for talent against higher-paying rivals without raising pay at all.

For career changers, the constraint is depth. General chatbot fluency is becoming table stakes; the differentiated skills — verification, prompt engineering, workflow design — are where the self-taught gap widens or closes. LinkedIn’s count of 1.3 million new AI jobs found most do not require a degree, which puts the emphasis on demonstrated capability rather than credentials — good news for self-teachers, provided they can prove it.

❓ FAQ

What did the iCIMS report find? 47 percent of US job seekers built AI skills in the past six months (up from 41 percent), self-teaching rose from 22 to 30 percent in a year, and employer-provided AI training stayed flat at roughly one in six workers.

Is AI training really a recruiting advantage? By the survey’s numbers, yes: 42 percent of job seekers said they would favour an employer offering AI training over a comparable one that does not, and 14 percent would accept lower pay for it.

Do these numbers apply to New Zealand? The survey is US-based, so treat it as a directional signal. NZ-specific data points the same way: KPMG’s finding that only 24 percent of New Zealanders have had any AI training, against near-universal use at work.

What AI skills are employers actually asking for? Lightcast posting data shows AI requirements spreading across industries, though AI-related roles remain a small share of total demand — 4 percent of US postings. Surveyed candidates mostly report general tool fluency, with far fewer holding specialised skills like prompt engineering or machine learning.

What should a worker do with this? If your employer will not train you, the market says build anyway — and prefer skills you can demonstrate (a portfolio, verified projects) over certificates, since recruiters “can no longer rely on simple resume keywords,” as iCIMS puts it.

🔍 THE BOTTOM LINE

The upskilling story of 2026 is not employers transforming their workforces. It is workers transforming themselves while employers stand still — and the employers who catch up will find the market has already priced training as a benefit worth pay. The 14 percent who would take lower pay for AI training are not desperate; they are signalling where they believe the returns are.