In 1930, John Maynard Keynes wrote an essay called “Economic Possibilities for Our Grandchildren” in which he predicted that within 100 years — by 2030 — technological progress would have solved the economic problem: the production of enough goods and services to meet human needs would require only 15 hours of work per week. The rest of time could be devoted to leisure, culture, and human flourishing.
We are not working 15-hour weeks. The standard workweek in most high-income countries remains 40 hours or more. Keynes was right about technological progress — productivity per hour worked has increased enormously — and wrong about where the gains went.
This history is worth keeping in mind as the current wave of AI-driven automation produces confident predictions about the future of work. The predictions are serious and the technology is real. Whether the social and economic consequences match the technological possibility depends on choices — about how productivity gains are distributed, how labour markets are regulated, what social contracts are maintained — that are political and institutional, not technological.
What is clear is that something qualitatively different is happening now. Previous waves of automation — mechanical, electrical, computerized — displaced physical and routine cognitive tasks while creating demand for non-routine cognitive work. AI is the first technology that can perform sophisticated non-routine cognitive work: writing, analysis, legal reasoning, medical diagnosis, creative design, code generation. The scale and scope of what AI can replace is larger than previous automation waves in ways that the standard “technology creates as many jobs as it destroys” reassurance may not adequately address.
What AI Can Actually Do Now
The capabilities of current AI systems are worth stating precisely, because they are often either overstated (general intelligence replacing all human work) or understated (mere pattern matching without genuine capability).
Language and text: Large language models can write fluently and coherently across genres, summarize complex documents, translate between languages with near-human quality, generate code from natural language descriptions, and converse in ways that users often find indistinguishable from human interaction. They can perform tasks that until recently required significant skill: drafting legal contracts, writing marketing copy, producing financial analysis narratives, generating news stories, composing academic essays.
Visual and creative: Image generation systems can produce original photorealistic images, illustrations, and graphic design from text descriptions. Video generation has advanced rapidly. Music composition AI is generating commercially viable content. These tools are displacing entry-level creative workers while also enabling new forms of creative production.
Code: AI coding assistants have measurably increased programmer productivity in controlled studies. GitHub Copilot, Claude, and GPT-4-based coding tools can generate working code for common tasks, debug existing code, and suggest implementations from high-level descriptions. Senior software engineers who use these tools effectively become substantially more productive; junior roles that primarily involve implementing well-defined specifications are most directly exposed.
Analysis and reasoning: AI can perform data analysis, identify patterns in large datasets, generate hypotheses, and reason through structured problems with a facility that compresses significant analyst work hours. McKinsey’s own analysis (with obvious self-interest disclosures) estimated that generative AI could automate 60-70% of the tasks currently performed by knowledge workers.
What AI cannot reliably do, currently, is: navigate genuine physical unpredictability (skilled trades, emergency response, caregiving), exercise moral judgment under conditions of genuine uncertainty, form genuine relationships and read subtle social dynamics, and exercise deep domain judgment in novel situations that diverge significantly from training distribution.
The Labour Market Effects: What the Data Shows
The economic debate about AI’s labour market effects is contested, and the evidence is still accumulating. But some patterns are already visible.
Wage polarization has accelerated: Labour markets in most high-income countries have polarized — high-skill, high-pay jobs growing, low-skill, low-pay service jobs growing, middle-skill routine jobs shrinking — since the 1990s, driven by previous waves of computerization. AI is accelerating this polarization and pushing it further up the skill distribution. Roles that were previously firmly in the middle-skill, middle-pay “safe” zone — paralegal, junior financial analyst, content creator, customer service representative — are now directly exposed.
Automation and wage effects are heterogeneous: The impact of AI automation on workers depends substantially on whether AI supplements human work (making workers more productive) or substitutes for it (replacing workers). The difference is partly technical — it depends on how tasks within jobs are structured — and partly a management decision. Companies that implement AI to augment worker capacity while reducing headcount produce different labour market outcomes than those that use AI to improve productivity without reducing headcount.
Creative and knowledge sector disruption is real: Graphic designers, stock image creators, junior copywriters, translators working on standard documents, and customer service agents have already experienced direct income effects from AI. The journalism industry has shed positions partly attributable to AI content generation. These are not speculative; they are documented current effects.
New jobs are being created: Prompt engineering, AI system maintenance, AI governance, AI training data curation, and AI-adjacent roles are growing. Whether these grow fast enough, at accessible enough skill levels, and with sufficient wage compensation to offset what is being displaced is empirically uncertain and depends on policy choices.
The Structural Question: Who Owns the Machine?
The most important political economy question about AI and work is not which jobs survive but who captures the productivity gains.
When AI makes a worker twice as productive, the conventional expectation is that the worker earns more — because their labour is more valuable. This is sometimes what happens, particularly for high-skill workers whose bargaining power allows them to capture productivity gains through higher wages or reduced hours.
More commonly, particularly in lower-skill roles and non-unionized workplaces, productivity gains from AI accrue to capital owners — shareholders, company owners — rather than workers. When a customer service team is reduced from 100 agents to 50 by AI tools that each agent 50% more productive, the 50 remaining agents do not automatically earn twice as much. The employer captures the cost reduction as profit.
The distributional question — who owns the AI and who captures its output — is not determined by technology. It is determined by bargaining power, labour market institutions, tax policy, and the degree to which workers have equity stakes in the productivity gains generated by their labour in combination with AI.
The history of previous automation waves does not provide unambiguous reassurance. Real wages in many high-income countries have stagnated or declined in real terms for decades while productivity has risen — with the gains captured primarily by capital owners. AI, if it produces the productivity gains that proponents project, could either reverse this trend (if workers share in gains) or accelerate it (if capital captures gains while labour is displaced).
Toward a New Employment Contract
The employment contract that developed over the 20th century — stable, long-term employment with benefit provisions, social insurance tied to employment status, pension accumulation through decades of service — was built for an economy of mass industrial employment with relatively slow skill change.
That contract is already under pressure from gig economy fragmentation, global supply chain restructuring, and the declining power of organized labour. AI intensifies every dimension of that pressure.
Several elements of a new employment contract are under discussion.
Portable benefits: Benefits including healthcare, pension contributions, paid leave, and accident insurance that follow individual workers regardless of employment relationship — eliminating the dependency on a single long-term employer that makes the existing system fragile.
Working time reduction: The historical response to productivity gains that exceed labour demand has been reduction in working hours — from 60-hour weeks to 48 to 40 over the 20th century. A genuine 30- or 32-hour standard workweek, with full pay maintained through productivity gains, is one mechanism for distributing AI-driven productivity gains through time rather than wages.
Worker equity: Profit-sharing, employee stock ownership, and co-determination mechanisms that give workers a stake in the productivity of the enterprises they work in, rather than only a wage claim on the labour they provide.
AI dividend: More radical proposals — associated with economists like Daron Acemoglu and philosophers like Nick Bostrom — suggest taxing the productivity gains from AI and using the revenue to fund social dividends, retraining programs, or expanded public goods provision.
The future of work will not be written by the technology. It will be written by the political choices societies make about how to govern it — who benefits, who is protected, what social obligations accompany the deployment of systems that eliminate human roles. Keynes’s vision of a 15-hour workweek was not wrong about what technology could make possible. It was wrong about what institutions would ensure.
Getting the institutions right this time is the most important economic policy challenge of the coming decade.
Also explore:
AI Job Displacement: The Future of Work
Universal Basic Income: Dream or Reality?
The Gig Economy: Digital Liberation or New-Age Exploitation? (published July 17)
The Great Wealth Transfer: $84 Trillion and Inequality (published August 27)
Related Articles
- Cryptocurrency Decentralization vs. Government Control: The Future of Your Freedom
- Universal Basic Income: The Future of Work Amid Automation
- Universal Basic Income Feasibility: Dream or Reality? The Global Debate
Weekly Geopolitical Briefing
Go Beyond the Headlines
Analysis of the forces shaping the world - delivered every week. No noise, no bias, just depth.
Subscribe Free ?No spam. Unsubscribe anytime.
About the Author
António Monteiro
Engineer by profession, geopolitical analyst by conviction. I believe responsibility for the planet's future doesn't belong only to governments and institutions - it belongs to all of us. Knowledge about geopolitics, international conflicts, and the forces shaping the world is the most powerful tool for becoming more conscious, informed citizens. You don't need to be a diplomat to understand what's at stake - you just need to want to go beyond the headlines. At Outside The Case, I analyze conflicts, power dynamics, and global trends with rigor and accessible language, so you can understand what's really happening in the world.
Read more about the author →