The Learning Tax: Why 76% of Workers Stopped Waiting

TL;DR

• 76% of workers are now using AI tools they signed up for themselves, not what their employer provided — because in 41% of cases, the employer provided nothing at all (Resume Now BYO AI Report, June 2026).
• A second framing surfaced almost simultaneously from Tel Aviv: the learning tax — the unpaid private hours workers now spend keeping up. Same behaviour. Opposite story.
• The pattern is global. US, Israel, India, Saudi Arabia, Canada, UK, Singapore all printed versions of the same story this week. India ranks #1 in AI economic capacity and #74 in workforce readiness — the widest gap in the QS Index.
• Two workers can spend identical hours on this and end up in opposite places. The difference isn't talent. It's whether the learning is a system or a debt.
• The honest action: stop waiting for training that isn't coming. Build a tiny weekly system you own. Make it visible.

The frame

There's a quiet identity decision sitting under every "I should learn AI" thought, and most people skip past it.

The decision isn't what tool to learn. It's whose job is it to learn it.

For thirty years, the answer was clear. Your employer paid for the training. They sent you on the course. The skill belonged to the company. You showed up and used it. If the world changed, they'd retrain you, eventually, probably.

That bargain just broke. Not slowly. This week.

Resume Now's BYO AI Report, published on 26 June and picked up by Forbes, GovTech, and Business Insider over the following 48 hours, surveyed 1,000 employed US workers and found something that ought to be on the front page of every HR magazine: 41% say their employer has provided nothing — no tools, no training, no guidance — to prepare them for AI at work. And 76% have already signed up for AI tools on their own and used them on the job anyway.¹

Three in four workers have decided not to wait.

That is not a productivity statistic. That is an identity statistic.

The two stories underneath the same number

Here is where it gets interesting. The same behaviour is being told two completely different ways, in two different parts of the world, in the same week.

In Tel Aviv on 23 June, the Jerusalem Post published a piece called The Rat Race 2.0. It introduced a phrase that is going to stick: "the learning tax."² The argument: employees now spend nights and weekends learning Cursor, prompting, new agent frameworks, the LinkedIn-mandatory tool of the month. Employers want AI-fluent workers. They don't give them the time, the budget, or the structure to become one. So workers pay the tax in private hours. Resentfully. Quietly. Permanently.

In Boston on 26 June, BCG's Julia Dhar — co-founder of BCG's Behavioral Science Lab — described the same workers to Business Insider, but called them something else: people with a "high-agency mindset."³ BCG's 2026 AI at Work research found 72% of workers saying skill expectations in their role have already changed, and 88% expecting major upskilling within five years. The ones thriving, Dhar said, are the ones who "show up with a belief that their actions will have an impact" — who treat the change as theirs to navigate.

Two framings. Same hours.

One is exploitation. One is compounding.

The interesting question isn't which is true. They're both true. The interesting question is: what makes the difference, for any given worker, between paying a tax and earning interest?

The evidence so far points to one variable, and it isn't talent or hours or industry. It's whether there's a system underneath.

The pattern, on five continents

This is not an American story. It's the closest thing to a synchronised global pattern we've had in workforce data this year.

Pick any continent. The story is the same. The skill expectations changed faster than any training program could keep up, and workers noticed before institutions did.

The India number is the one to sit with. Number one in the world for AI economic capacity. Number seventy-four for workforce readiness. That gap, played out across a billion working-age people, is the largest single-country learning tax on the planet.

What this isn't

A few things this story is not about, which the hype layer keeps confusing.

It isn't really about AI tools. Cursor, ChatGPT, Claude, NotebookLM — they'll be different by Christmas. The behaviour underneath is what matters: workers deciding to take responsibility for staying employable when their employer didn't.

It isn't about Gen Z hustle culture. The BCG data covers all working ages. The Resume Now sample is the full US workforce. The Jerusalem Post framing skews older if anything. This is not a generational moment.

It isn't about "side hustles" or career change. Most of these workers aren't leaving. They're trying to stay valuable in the job they already have.

And it isn't, despite a thousand LinkedIn posts insisting otherwise, a moral story. The workers paying the learning tax with resentment aren't lazy. The workers compounding it with intent aren't superior. They're often doing identical hours. The system is what's different.

The mechanism — why tiny systems beat heroic effort

Here's the part worth being concrete about, because the data is unusually clear.

Two workers. Same job. Same week. Both spend roughly four hours of private time on AI learning.

Worker A opens YouTube on Sunday night, watches whatever's trending, tries three tools, abandons two, can't remember on Thursday what she learned. By month four, she resents the time. By month eight, she stops doing it. By month twelve, she's behind.

Worker B picks one tool every two weeks. Spends 30 minutes a day, three days a week. Writes one paragraph after each session: what I tried, what worked, what I'll use it for at work tomorrow. By month four, she has a notebook. By month eight, she has a public artefact — a doc, a Loom, a small internal training. By month twelve, she's the person her team asks.

Same hours. Different identity.

The atomic unit isn't the hour. It's the loop: pick one thing, practise it small, write down what you learned, use it visibly within seven days.

This is the part that compounds, and the part that the people writing think-pieces about "the AI revolution" mostly skip. The revolution isn't the tool. It's whether you can build a personal learning system that survives a bad week.

What this means for you

Recommendations addressed to anyone with a job that AI is touching — which, in 2026, is approximately every job.

If you've been waiting for your employer to train you. They are statistically unlikely to. Forty-one percent provide nothing. Of the rest, most provide a single town-hall and a Slack channel. Stop waiting. Pick one tool — the one your team will actually use, not the one Twitter is excited about this week — and give it 30 minutes, three times a week, for six weeks. That is enough to be the second-most-fluent person in most rooms.

If you've been learning, but resentfully. This is the learning tax framing — and it's emotionally accurate but professionally expensive. The hours go in either way. The question is whether they leave a trail. After each session, write three lines: what you tried, what worked, what you'll use at work this week. This converts the tax into evidence. Evidence is what gets promoted, paid, or recruited.

If you're a manager. Your team is already paying for AI training. You're just not paying for it — they are, in private hours, with no structure. The cheapest, highest-leverage intervention available to you in Q3 is one hour of paid Friday "learning time" per fortnight, with a five-minute show-and-tell at the end. This costs almost nothing. It externalises the learning tax onto the company where it belongs, and it gives you visibility into what your team is actually using. Skip the LMS procurement cycle. Start next Friday.

If you're in HR or L&D. The Resume Now data is your business case. 41% of workers report no AI training; 76% are using consumer tools on company devices anyway. The shadow-AI security risk is real, the productivity gain is real, and both are happening without your involvement. A six-week "AI Foundations" sprint that names two approved tools, two banned ones, and one weekly practice ritual will outperform any 12-month strategic plan you commission this year.

If you're a policymaker or educator. The QS World Future Skills Index 2027 number for India — #1 in economic capacity, #74 in workforce readiness — is the largest gap of its kind ever recorded in an index of this type. The decision now is whether to close it through national skilling programmes or let private actors capture the rent. Singapore is closing it institutionally. India is, for now, letting individuals pay the tax themselves.

What we don't know yet

• Whether the BYO AI behaviour persists, or collapses into resentment. The Jerusalem Post framing predicts collapse within 12–18 months for workers without structural support. The BCG framing predicts compounding gains. Both can't be right at population scale; one will dominate.
• Whether employers respond before the talent leaves. Robert Half's 44% job-search intent suggests the window is short. Canadian data tends to lead other markets by 6–9 months on workforce sentiment.
• Whether the India gap closes nationally or unevenly. The Fukuoka-Haryana skilling agreement signed this week is one signal of bilateral, sector-specific solutions outpacing multilateral ones.
• Whether "the learning tax" becomes a labour-organising frame. It has the linguistic shape of one. Tipped wages became a useful phrase the moment someone said it out loud.

Bottom line

The workers compounding right now did not get more time, more tools, or more talent. They got tired of waiting. They picked one tool, built a thirty-minute weekly ritual, and started writing down what they learned. That's the whole technique. The interesting part isn't the AI — it's the moment they decided the learning was theirs to own. The workers who haven't decided yet aren't behind because of skill. They're behind because they're still waiting for someone to tell them it's time to start.

Sources

Tier 1

• Forbes, The Rise Of 'Bring Your Own AI' To Work As Leaders Fall Behind, 28 June 2026
• Business Insider, Workers with this mindset are thriving in the AI era, says BCG leader, 26 June 2026
• The Hindustan Times, Japan turns to Haryana to fill 50k skilled workforce vacancies, 30 June 2026
• The Times of India, India Number 1 in economic capacity, 74th in workforce readiness: QS, 28 June 2026
• The Washington Post / The Hill, retirement & workforce purpose coverage, 15 June 2026

Tier 2

• Resume Now, BYO AI Report, June 2026 (primary data source: n=1,000 US workers)
• GovTech, AI at Work: Employees Aren't Waiting for Permission, 28 June 2026
• Jerusalem Post, The Rat Race 2.0: The less glamorous side of the AI revolution, 23 June 2026
• Insurance Times, Rising Professionals' Forum 2026 coverage, 25–29 June 2026
• Consulting.ca / Robert Half, More than 4 in 10 professionals plan to look for a new job in H2 2026, 16 June 2026
• TradeArabia / SAP–YouGov, Saudi firms see strong AI returns amid workforce overhaul, 24 June 2026
• QS World Future Skills Index 2027, released 25 June 2026
• BCG, 2026 AI at Work Research (cited via BI)
• Media OutReach / SIM GE Singapore, 24 June 2026

Footnotes

1. Resume Now, BYO AI Report, June 2026. Sample: 1,001 employed US workers, surveyed Q2 2026. Reported via Forbes (28 June) and GovTech (28 June).

2. Asaf Ronel, The Rat Race 2.0: The less glamorous side of the AI revolution, Jerusalem Post, 23 June 2026.

3. Julia Dhar in Business Insider, 26 June 2026, drawing on BCG's 2026 AI at Work Research (n undisclosed at time of writing; corporate sample).