# Most of us are learning this in our own time

> Most of what I know about working with AI I learned in the evenings. Nobody asked me to and nobody paid for the hours. I tried things, got them wrong, and asked people who were a little further along, and I enjoyed most

- Source: https://www.aicapability.academy/insights/capability-edge-edition-16
- Author: Marco van Hout, AI Capability Academy (Demadera)
- Published: 2026-09-29
- Type: newsletter edition (The Capability Edge)

Most of what I know about working with AI I learned in the evenings. Nobody asked me to and nobody paid for the hours. I tried things, got them wrong, and asked people who were a little further along, and I enjoyed most of it. Then last week I read that 56% of employees at large companies get no time at work to build these skills, and I started to wonder who that way of learning works for, and who it leaves behind.

A few of the things I read last week:

- In a survey of 1,000 salaried employees at US companies with more than 5,000 staff, 56% said they have no time in their working hours to build AI skills, and nearly 60% said AI skills do not count in promotion decisions. (Workera)
- Abhinav Bohra, a senior applied scientist at Amazon, spends 8 to 12 hours a week and around $3,000 a year learning AI outside work, and describes that learning as part of the job "even when it happens outside the job". (Business Insider)
- Of more than 1,000 US workers surveyed, 44% feel overwhelmed by the pressure to learn AI, and 30% say they are likely to pay for AI training out of their own pocket. (Resume Now)
- A UK government review found that people mostly learn AI "through trial and error, peer support, online videos, or vendor-led prompts", and that women are more likely than men to have had fewer chances to build digital skills. (Skills England and DWP)
- In occupations mostly held by women, AI exposure is spread evenly across every wage level, while in occupations mostly held by men it sits mainly in the better-paid roles. (Fernandez et al., arXiv)

So we are asking people to learn one of the biggest changes in their working lives on their own time. What does that do to who gets to keep up?

## Evenings are not evenly shared

I was lucky. I had evenings to spare, work that rewards curiosity, and friends who knew more than I did. But a way of learning that runs on free evenings only works for people who have them. The person caring for a parent, the nurse on nights, the parent on a part-time contract: they meet the same tools at work with far less room to practise.

The Fernandez paper made this concrete for me. In jobs mostly held by women, AI reaches the lower-paid roles as much as the better-paid ones. The UK review says women are also more likely to have had fewer chances to build digital skills in the first place. When I put those two findings next to each other, leaving the learning to each individual stops looking neutral. It gives the most practice to the people who already had the most. As we shared on Wednesday, the Konstanz team in Germany found the same shape by education, and Elena Gerdiken said it plainly: the better educated, in knowledge work and in larger organisations, "are currently benefiting most".

The Workera promotion number worries me too. If people learn at night and it does not count at review time, we are asking them to invest in something the organisation says it values and then chooses not to see.

## Where the learning should happen

The strongest case against me comes from Ethan Mollick, who has argued for a while that the best uses of AI are found by the people doing the work. People who know their job well can tell through trial and error when AI helps, he writes, "in the way that outsiders (and even AI-savvy junior workers) cannot." I agree with him. Self-directed learning is not the problem, and a lot of the good practice I have seen started with one curious person. The question for me is where and when that learning happens. Something discovered during the working day, with colleagues around, can become something a team shares. Something discovered at the kitchen table stays with one person, and it depends on that person having an evening to spend.

Some organisations have tried to move the learning back into the day. In 2025 Canva gave all of its more than 5,000 employees a week away from their normal work to learn AI together, role by role; SaaStr estimated the payroll cost alone at around $10 million. Not every organisation can do that, and I do not think it has to. In a Study.com survey reported in July, 46% of workers said they were self-taught and 33% had received formal training, and 67% said two hours a week or less would be enough to build the skills. Trevor Schulze, CIO at Genesys, named the gap when he warned against "declaring AI a priority without creating any structural space for learning."

Learning in the working day also means learning beside someone, and that is where people pick up the habit of checking a model's output instead of only producing it.

*If your team wants a simple way to practise that checking together, the Verify playbook was made for exactly that.*

## If you work in L&D, HR, or transformation

We usually know who is learning in their own time, because they are the ones who come to us with questions, and I understand the pull to build on their enthusiasm. Two things worth trying this week. Ask one team how many hours each person spent learning AI outside work last month, without any judgement, and then ask who could not spend any. That second list is where your programme should start. Then put two protected hours a week in the calendar for one group, ideally lower-paid or part-time staff, and let them spend those hours on their own work with someone a little further along sitting beside them.

*If you would like to set that up alongside people doing the same in their own organisations, that is what the AI Capability Academy programme at aicapability.academy is for.*

## The provocation

Think of the person on your team with the least spare time outside work. How are they learning AI right now, and what would it take to give them two hours of it inside the working week, starting this week?

## Sources

- [Workera, AI Training More Than Doubled This Year, but 56% of Employees Report No Time at Work to Build the Skills (23 September 2026)](https://www.prnewswire.com/news-releases/ai-training-more-than-doubled-this-year-but-56-of-employees-report-no-time-at-work-to-build-the-skills-workera-research-finds-302887120.html)
- [Jacob Zinkula, Tech workers are spending nights and weekends learning new AI tools. They say they can't afford not to, Business Insider via Yahoo Finance (21 June 2026)](https://finance.yahoo.com/technology/ai/articles/tech-workers-spending-nights-weekends-110001055.html)
- [Resume Now, AI-whelmed Worker Report, via Allwork.Space, 44% Of Workers Are 'AI-Whelmed' As Pressure To Learn AI Grows (15 September 2026)](https://allwork.space/2026/09/44-of-workers-are-ai-whelmed-as-pressure-to-learn-ai-grows)
- [Skills England & Department for Work and Pensions, What works for AI upskilling in the UK: insight briefing based on research (10 June 2026)](https://www.gov.uk/government/publications/skills-for-ai-what-works-for-ai-upskilling-in-the-uk/insight-briefing-based-on-research-what-works-for-ai-upskilling-in-the-uk)
- [Skills England & Department for Work and Pensions, What works for AI upskilling in the UK: research evidence, analysis and methodology (10 June 2026)](https://www.gov.uk/government/publications/skills-for-ai-what-works-for-ai-upskilling-in-the-uk/research-evidence-analysis-and-methodology-what-works-for-ai-upskilling-in-the-uk)
- [Miriam Fernandez, Ángel Pavón Pérez et al., When AI Enters the Workplace, Who Faces Greater Risks? A Gendered Analysis, arXiv 2609.21756 (18 September 2026)](https://arxiv.org/abs/2609.21756)
- [University of Konstanz, Future of Work Lab (Kunze, Opitz, Gerdiken), The use of AI at work: slow, informal and unequal, via EurekAlert (September 2026)](https://www.eurekalert.org/news-releases/1144563)
- [Ethan Mollick, Making AI Work: Leadership, Lab, and Crowd, One Useful Thing (22 May 2025)](https://www.oneusefulthing.org/p/making-ai-work-leadership-lab-and)
- [SaaStr, Canva gives its 5k employees a week to learn AI. Will it work? (8 July 2025)](https://saastr.com/canva-gives-its-5k-employees-a-week-to-learn-ai-will-it-work/)
- [Alison Roller, Study finds two hours a week could close the AI skills gap, TechTarget (22 July 2026)](https://www.techtarget.com/searchcio/feature/Two-hours-a-week-could-close-the-AI-skills-gap)