Work-Life Balance Is Dead. Meet AI-Human Balance.

Work-life balance has been pushed aside. The negotiation that defined three decades of working life quietly settled itself the day AI showed up. And the terms should make every executive uncomfortable.

As AI is adopted in the workplace, many employees have ironically found greater balance, leveraging the efficiency of AI to buy back time for personal interests and pleasures. The ROI of AI has gone to the beach. I mean that literally. A Bank of Korea study of more than 5,500 workers found generative AI cut working hours by nearly four percent — roughly ninety minutes a week — with essentially no corresponding rise in output. The researchers' explanation: workers absorbed the efficiency as leisure on the job. A Stanford study tracking 200,000 American households found the same pattern at home — the hours ChatGPT saves go to leisure, not learning. Stanford's own headline said the productivity boost is happening from the sofa.

The efficiency gains have been absorbed by workers doing less, not by improving overall corporate efficiency. MIT's Project NANDA research found roughly 95% of the enterprise AI initiatives it examined were producing no measurable P&L impact — while workplace AI use has doubled since 2023. Consumer tracking from Occam, Alpharoc's verified human-data research system, shows the same settling-in: the share of people using AI multiple times a day has climbed steadily for two years, and the "never" users have shrunk from four in ten to fewer than three. Adoption is settled. Value capture isn't. The gains are real; they're just leaking everywhere except the P&L.

And while that ledger goes unaudited, we continue to debate the role of AI and the role of humans. I think the debate is asking the wrong question. Here's the thing nobody wants to hear: the imbalance isn't between AI and jobs. It's between what we delegate to machines and what we keep for ourselves. And right now, almost nobody is choosing deliberately.

Machine confidence, shipped without judgment

We recognize that AI without humans can create misguidance and misleading information. What's remarkable is who keeps proving it. Deloitte Australia agreed to refund the final installment on a A$440,000 government contract after fabricated academic references and an invented court quotation were found in an AI-assisted report. EY Canada withdrew a study whose sources included a McKinsey report that does not exist. A global law firm apologized to a New York court for an AI-assisted filing riddled with inaccurate citations. One legal researcher's database of AI-fabricated citations caught by judges shows the curve: ten rulings in 2023, thirty-seven in 2024, seventy-three in the first five months of 2025 — increasingly committed by qualified professionals, not amateurs. These are organizations built on verification, shipping machine confidence without human judgment.

The public, interestingly, is ahead of the enterprise here. Occam's consumer research — built on Alpharoc's principle that humanity is best informed by human data — found nearly six in ten consumers who use AI to find something to buy spend time verifying what it tells them. More than half of those verifiers go straight back to Google to check. Nobody mandated this. Ordinary people built their own human verification layer on top of the machine, while corporations are still shipping unreviewed reports. Some consumers are practicing a dimension of AI-human balance instinctively.

Generation is cheap. Taste isn't.

We also see that whilst AI can generate content endlessly, it doesn't have the human senses of taste and curation. Watch what happens when judgment gets automated out of the loop. Late last year, Meta's ad system auto-replaced a menswear brand's best-performing creative — a model matched precisely to its audience — with an AI-generated image so viscerally off-brand it went viral for the wrong reasons. The technology worked exactly as designed. The taste was missing. A creative director I'd happily quote in any boardroom put the distinction perfectly: "Prompting is a tactic. Direction is a discipline."

“AI executes. Humans judge. The brands that win this decade are the ones who get that distinction right.”

The displacement debate is the constant

The debate about how many people will be displaced by AI continues, and it is genuinely unresolved. Anthropic's CEO Dario Amodei warns of a possible "white-collar bloodbath"; the IMF's Kristalina Georgieva says AI is hitting labor markets "like a tsunami." Meanwhile the World Economic Forum projects that technology and other macrotrends will displace 92 million jobs by 2030 while creating 170 million — a net gain of 78 million. Both camps hold serious evidence. Like all major evolutions in human practice, technology, industry, and manufacturing, the argument will continue until the next major shift arrives. That's not a reason to ignore it. It's a reason to stop waiting for the debate to resolve before managing the balance yourself.

Slop: the feeling, made official

The addictive nature of AI is creating a world where we are becoming over-reliant and over-dependent, producing excessive content and slop. Don't take my word for it, take the dictionary's. Merriam-Webster's 2025 Word of the Year is "slop": low-quality digital content mass-produced by AI. Australia's Macquarie Dictionary independently chose "AI slop." Collins picked "vibe coding." When three dictionaries crown AI-junk vocabulary in the same year, the culture has rendered its verdict.

The workplace version now has a name and a price tag. Researchers at Stanford and BetterUp call it "workslop" — AI output that masquerades as good work but lacks the substance to advance the task. Forty percent of desk workers received some in the past month. Each piece takes nearly two hours to untangle, an invisible tax of about $186 per employee per month — over $9 million a year for a 10,000-person company. And it corrodes trust: half of recipients rate the sender as less capable afterward. We often feel that using AI to produce more content and feed more information to others is effective. Step back and the evidence is blunt: it isn't. There's an individual cost too. The same MIT study measured brain activity while people wrote with ChatGPT and found the weakest neural connectivity of any group. Participants couldn't quote their own essays minutes after writing them. The researchers coined a term I suspect we'll be using for years: cognitive debt.

The creation cycle

With AI tools and apps being developed at the fastest rate ever, we are caught in a creation cycle that has us glued to our screens like never before. The New York Times reported this month on what's happening inside Apple's App Store, and the numbers tell the whole story. New app releases, which had declined for nearly a decade, jumped 30 percent last year to about 600,000 and then roughly 560,000 more landed in just the first half of this year, nearly doubling the pace. The driver is "vibecoding": describing an app in plain English and letting AI write the code. Now the downside. While the supply of apps exploded, downloads barely moved — up just 2 percent. The machines are building; the humans aren't downloading. A former head of the App Store described the incoming wave to the Times as "stuff that somebody might use once." We have democratized creation without democratizing judgment about what's worth creating.

The snake eats its tail

Every piece of content we create with AI feeds the internet — which is the fuel station for AI. This cycle will spin continuously, degrading the quality of content over time as the system feeds off itself without human intervention. That is no longer speculation. Oxford-led researchers demonstrated it in Nature: they call it "model collapse." Train successive generations of AI on indiscriminately recycled machine output and the models progressively lose the tails of the original data — like photocopying a photocopy until the image dissolves. The fuel is already diluted: one Ahrefs analysis found nearly three-quarters of newly published English-language pages contained at least some AI-generated material, most of it a human-machine blend. Follow that to its conclusion and you arrive somewhere important: genuine human-produced thinking becomes the scarce, precious resource. That is the economic case for the human half of the balance, not nostalgia. Scarcity.

Meanwhile, it is getting harder and harder for the AI-dependent to identify quality and reality. In one industry study, just 0.1% of two thousand participants could spot every deepfake shown to them even when told to look. AI-generated faces are now rated as more trustworthy than real ones. Most people say telling human from machine matters enormously; most admit they can't reliably do it. What fills that gap is the thing no model can generate: real-life experience and the wisdom accumulated over many years of living. Judgment built on experience is becoming the last reliable detector.

Naming the balance

My overall thesis is that we need to start thinking about AI-human balance the way we once thought about work-life balance. Something you manage deliberately, with boundaries and choices, rather than something that happens to you. And my definition of the human side goes beyond a person and their time engaging with AI. It's the impact on humanity of AI. The two are interconnected at every level: the individual (what cognitive debt does to your own thinking), the employee (whose saved time went to building sand castles), the organization (whose AI investment shows no return while workslop taxes every team), the community and nation (where trust in what we read is eroding), and the species (whose collective information commons is beginning to feed on itself).

The public already senses it. Pew's latest research found half of Americans are now more concerned than excited about AI in daily life, yet nearly three-quarters would still happily let AI help with everyday tasks, and six in ten want more control over how it's used in their lives. Read those numbers together and the conclusion is unavoidable: we want the help, and we fear the cost. That is not an adoption problem. That is a balance problem. Work-life balance took a generation of deliberate effort — boundaries, rituals, and hard conversations — before it became something organizations managed on purpose. AI-human balance deserves the same discipline, starting now, while the choices are still ours to make.

So here's my question, and I genuinely want your answer: where in your business does a machine now decide something a human used to judge — and did anyone choose that on purpose?

Postscript — This paper was created with the help of my brain, the internet and AI working together. I created the concept, the key points based on extensive reading over time, drafted the headlines, key points and messages. I used the internet and AI for research, research audits and verifications. I used AI to draft. I edited every word. I applied my taste and a dose of trust.

---

## Sources

1. Bank of Korea, BOK Issue Note: Generative AI and Labor Market Outcomes (Suh, Oh & Yoon), English edition, June 9, 2026. Survey of 5,512 Korean workers.

2. Stanford working paper on generative AI and household time use, tracking browsing data from more than 200,000 U.S. households; summarized in Stanford Report, April 2026.

3. MIT Project NANDA, "The GenAI Divide" (2025), MIT Media Lab; Gallup workplace AI tracking, 2023–2025.

4. Occam consumer tracking (Alpharoc), monthly panel May 2024 – June 2026, n=5,313.

5. Reporting on Deloitte Australia's AI-assisted report for the Department of Employment and Workplace Relations, October 2025; EY Canada study withdrawal, May 2026.

6. Damien Charlotin's database of court rulings involving AI-fabricated citations, as reported by Business Insider, 2025.

7. Occam consumer survey (Alpharoc), March 2026, n=1,204 U.S. adults.

8. Grace Liu, quoted in Forbes, "AI Isn't Replacing Creativity. It's Moving It Upstream," May 26, 2026. Advertising case: Forbes, February 13, 2026.

9. Dario Amodei, interview with Axios, May 28, 2025; Kristalina Georgieva, remarks on AI and labor markets, February 2026.

10. World Economic Forum, Future of Jobs Report 2025 (survey of 1,000+ employers representing 14 million workers across 55 economies).

11. Merriam-Webster, Macquarie Dictionary, and Collins Words of the Year 2025.

12. BetterUp Labs & Stanford Social Media Lab, "AI-Generated 'Workslop' Is Destroying Productivity," Harvard Business Review, September 2025 (1,150 U.S. desk workers).

13. Kosmyna et al., "Your Brain on ChatGPT: Accumulation of Cognitive Debt…," MIT Media Lab, 2025 (arXiv:2506.08872).

14. Kalley Huang, "A.I. and the App Store," The New York Times, July 2026, citing Sensor Tower data.

15. Shumailov et al., "AI models collapse when trained on recursively generated data," Nature 631, 755–759 (2024); Ahrefs content analysis of 900,000 pages, April 2025.

16. iProov deepfake detection study, 2025 (n=2,000 UK/US adults); Nightingale & Farid, PNAS (2022); Pew Research Center, 2025.

17. Pew Research Center, "How Americans View AI and Its Impact on People and Society," September 17, 2025 (5,023 U.S. adults).