“The budget for training is zero, but it’s not zero.”

That is what a product manager at a North Carolina tech company was told when she asked for help learning AI. She works closely with engineers who already use coding assistants every day, while her own role came with no tools and no plan. She could feel the gap widening. “I have to learn it on my own,” she told me. So she went to leadership and asked, in her words, whether there was an appetite for training her. The answer was a shrug in corporate grammar: not a yes, but not a no.

When a company carries the adaptation, learning AI is a task at work. When it will not, the learning moves home, onto the worker’s own time, at her family’s expense.

This spring I interviewed working mothers across North Carolina about how AI is changing their jobs. I expected the differences to fall along skills or industries. The sharpest difference was simpler. It was whether the employer took the lead. When a company carries the adaptation, learning AI is a task at work. When it will not, the learning moves home, onto the worker’s own time, at her family’s expense.

I saw what carrying it looks like. Another interviewee’s company treats AI training the way it treats any core system. The mandatory sessions come before the license, same as Salesforce. They brought in a trainer from OpenAI and ran a 101, a 201, and a 301. The day before we spoke, roughly 600 employees sat in on the training call, one of several sessions run across time zones. She found the 101 too basic and moved herself up to the 201. That is the point. A floor is being raised under everyone, including the colleague who had never opened the tool at all. Her learning happened on the clock, with the company footing the bill.

Now go back to the product manager. She did the training department’s job herself. She asked the free version of ChatGPT which programs in her field were most respected, worked through what it gave her, and built a shortlist of six options. Nobody assigned this. Nobody is paying for it. She is also raising three kids under seven. The hours this takes have to come from somewhere, and none of the somewheres belong to her employer.

We file AI under workplace technology, a question of tools and training budgets. From the kitchen table it reads differently. It is a question of whose hours absorb the transition. Hours the company pays for, or hours a family loses. The technology is identical in both cases. The only thing that moves is the cost.

And something else moves with it. The woman whose company trains her learns the tool, and underneath it, that she is worth investing in. The woman left alone becomes an unpaid student of her own profession, and the lesson underneath settles in: keeping up is your problem. One company put 600 seats on a call to carry its people through the transition. The other looked at its product manager and said zero, but not zero.

So if you want to know what AI is actually doing to work, do not only ask the companies counting their savings. Ask the mother assembling her own curriculum with a chatbot because nobody at her job would assign one. She can tell you exactly what an unsupported transition costs. She is the one paying.

Logan Currie is a Visiting Fellow at Capita’s Family Policy Lab.