AI-generated article. This article was researched and drafted using AI tools and published automatically, and its featured image was generated by AI. Facts are drawn from the sources cited in the text.
Ask a woman running a small business which of her roles AI will touch first, and she will usually name the ones closest to her own desk. Scheduling. Invoicing. The inbox. Answering the same customer question for the ninth time in a week. That instinct turns out to be correct. But the AI job growth gap now showing up in US employment data is not mainly a story about jobs disappearing. It is a story about who is being hired into the work AI touches.
What Pew actually measured
A Pew Research Center analysis published on October 22, 2026 scored 480 occupations for generative AI exposure using Department of Labor O*NET ratings, then compared employment growth in 2023 to 2026 against 2017 to 2019, the years before chatbots were widely available. For every one point increase in AI exposure, men’s employment grew 6.3%. Women’s moved 0.2%. At the very top of the exposure scale, men’s employment was up 17.8% and women’s up 0.5%.
Read the other end of the scale and it reverses. In the least exposed occupations, women’s employment grew faster than men’s. The divergence only opens as exposure rises, which is what makes it hard to wave away.
Where the AI job growth gap turns sharp
The split runs along education. Among workers with a high school diploma or less, Pew found men’s employment rose 8.8% per exposure point while women’s fell 9.5%. With some college, men gained 3.6% and women lost 6%. Among degree holders, no significant difference at all.
Pew does not claim exposure caused this, and that caution is worth keeping. The likeliest explanation on offer is plain occupational concentration. As Forbes summarized the findings, men without degrees cluster in truck driving, construction labor and freight handling, near the bottom of the exposure scale. Women without degrees cluster in secretarial, reception and customer service work, which sits in the middle to upper part of it. The same roles behind the squeeze on administrative jobs are the ones the growth data has now caught up with.
This is a structural gap, not a confidence problem
A Harvard Kennedy School Shorenstein Center study released on October 14, 2026, surveying more than 5,000 US adults, found women were likelier than men to report being left out of decisions about AI, and to say they lacked access to training on how AI is used at work. The same research counted well over 100,000 experts quoted in US AI coverage and found men appearing 4.3 times more often than women.
The workplace numbers point the same way. McKinsey and LeanIn data cited by Forbes found 33% of men at entry level are encouraged by a manager to use AI tools, against 21% of women. Moving out of an exposed role is hard, too. Only around 14% of occupational transitions are direct skill transfers, men are roughly 30% likelier to have one of those pathways, and 85% of new jobs require a credential. Pew’s June 2026 survey found 68% of women say AI is advancing too quickly, against 58% of men. Next to the growth data, that looks less like reluctance than accuracy.
The more useful question is not whether women will use AI. It is who gets paid to.
What it means for the woman who owns the business
An owner is not an employee, but the exposure map still applies to her own task list, and adoption is already tilted. JPMorganChase Institute research from May 2026, built on Chase business banking data, found 19.7% of male-owned firms had adopted AI by the end of 2025 against 17.2% of female-owned firms. The absolute gap widened from 0.3 points in 2019 to 2.5, more than eight times larger. The authors point to lower revenues, tighter capital, caregiving time, and fewer female-owned firms having employees at all (17% against 23.5%).
Here is the part that cuts both ways. The mid to high exposure tasks are exactly the ones a stretched owner can genuinely hand off, and that relief is real. They are also the roles she might have hired a person into one day, and employers are quietly not growing them. Both are true at once, which is why a straight “adopt faster” answer does not fit.
One small place to start
Write down the five tasks that ate the most hours last month. Mark the ones that are mostly moving text and data around, because that is roughly what the exposure score measures. Pick one. Spend thirty minutes testing whether a tool handles it at even 70% quality. Smaller than any training program, and it gives something no survey can: a first-hand read on where the line falls in this particular business. Starting at all puts an owner ahead of most, and the training gap data suggests the people handed that space are not the ones who need it most.
Pew says plainly it cannot yet tell whether the next decade brings net job loss or growth. What the data does say is that the growth so far has not been shared evenly, and that it thinned out most for the women with the fewest credentials to trade with. Whether that reads as a transition or a closing door depends on who is given space to learn in the next two years, and on who ends up doing the hiring. For a small business that is not an abstract policy question. It is the same decision as whether to open an entry-level role at all.
