AI Automation Is Coming — Do We Have a Plan for Workers?

The Real AI Race Isn't About Chips — It's About People

Everyone is talking about the AI race: who builds the best models, who controls the most powerful chips, who dominates the data centres. But there is a parallel competition happening that almost nobody in the technology industry is discussing with the same urgency — the race to keep societies functional as AI reshapes the workforce from the ground up.

Former US Secretary of Commerce and Rhode Island Governor Gina Raimondo put it starkly in a recent TED talk: if America wins the AI technology race but displaces tens of millions of workers in the process, it will have "automated its own decline." Recession, social unrest, and political instability don't just cause human suffering — they generate the kind of regulatory backlash that can strangle a technology industry in its infancy. The countries that manage this transition well won't just be more humane. They'll be more competitive.

This isn't a hypothetical problem neatly scheduled for the 2040s. It is happening now, and the systems we have to deal with it were built for a different world entirely.

Why Tens of Millions of Workers Are Already Exposed

Estimates vary, but researchers broadly agree that somewhere between 40 and 80 million American workers hold jobs that are significantly vulnerable to AI-driven automation in the next decade. These are not exclusively low-skill, low-wage positions. We are talking about accountants, paralegals, radiologists, software testers, customer service managers, data analysts, and junior financial advisors — the kinds of middle-income professional roles that were once considered safe because they required education and judgment.

What makes this wave different from previous technological disruptions — mechanisation, computerisation, offshoring — is its speed and breadth. When factories moved to Asia in the 1980s and 1990s, the disruption was brutal but geographically and sectorally concentrated. Rust Belt communities bore the weight while coastal knowledge workers largely carried on. AI does not work that way. It scales horizontally across industries simultaneously, and it is particularly adept at the cognitive, pattern-recognition tasks that make up a huge proportion of white-collar work.

Raimondo recalls watching her father lose his career at the Bulova watch factory in the early 1980s after almost 30 years of service. He was 56 with no bridge to a new chapter. The manufacturing exodus affected perhaps two or three million workers. The AI transition could affect an order of magnitude more, and if the policy response is equally inadequate, the political and social consequences will dwarf what followed deindustrialisation — which, Raimondo argues, we are arguably still living through in the form of polarised, dysfunctional politics.

The Two Wrong Answers — and What a Real Plan Looks Like

The public debate about AI and jobs has largely collapsed into two unhelpful camps. The first wants to slow or halt AI development through heavy regulation, treating the technology itself as the problem. The second proposes Universal Basic Income as a clean solution — give everyone a payment and let the market sort out meaning and purpose. Both miss the point.

Slowing AI doesn't work because the technology is global. Regulatory drag in the United States or Europe doesn't pause development in China — it simply shifts where the breakthroughs happen and who sets the norms. And UBI, while not without merit as a partial safety net, fundamentally misunderstands what work provides. A job is not just a paycheque. It is structure, identity, social connection, and purpose. Societies where large numbers of people are paid not to work don't flourish — they fragment.

A serious transition plan has to sit in a harder, less comfortable middle ground. It requires redesigning incentive structures so that companies find it cheaper to retrain workers than to lay them off. It requires workforce training systems built around employer needs rather than enrollment numbers. And it requires a career transition safety net designed for a world where people will change not just jobs but entire fields multiple times across a working life.

Breaking the Wall Between School and Work

One of the most durable and damaging assumptions embedded in Western education systems is that learning ends when employment begins. You go to school, you get a degree, you enter the workforce — and then you are largely on your own when that workforce changes under your feet.

This model made imperfect sense in an era of stable, decades-long careers. It makes no sense at all in an AI economy where job requirements will evolve continuously. The accountant who automates routine close-of-books work will need to develop advisory, interpretive, and relational skills that were previously optional. The software tester replaced by automated quality assurance tools needs to understand how to supervise and interrogate those tools. These transitions require real, structured learning — and they need to happen while people are still employed, not after they have already been let go.

Raimondo points to a model that actually works: the partnership built around TSMC's expansion into Arizona. The company specified exactly which skills it needed — electrical engineers, equipment operators with particular competencies — and then worked with community colleges and accelerated certificate programmes to train directly for those roles. The result is a functioning advanced chip manufacturing operation producing leading-edge AI hardware on American soil for the first time.

This is not a radical idea. Apprenticeships, co-operative education programmes, and employer-led training have existed for decades. The problem is scale and stigma. These routes cover a tiny fraction of the post-secondary system, and in a culture that equates a four-year university degree with success and everything else with failure, participation carries a social cost that puts many people off. Changing that will require changing funding mechanisms — schools and programmes should be funded based on employment outcomes, not enrollment — and changing cultural signals from political leaders and employers alike.

Modernising the Career Safety Net for AI-Era Disruption

America's unemployment insurance system was designed in a different century for a different economy. It assumes that job loss is temporary, that workers will return to roughly the same role, and that the primary need is short-term income replacement. In practice, it provides inadequate support for middle and higher earners, does nothing to fund retraining, and creates no pathway into self-employment or business creation.

A 21st-century career transition system needs several components that the current one lacks entirely. Temporary wage insurance — sometimes called wage top-up — could bridge the pay gap when a displaced worker takes a lower-paid entry-level position in a new field. Without it, the rational economic decision is to stay unemployed and job-hunt within a shrinking sector rather than accept a pay cut to start over somewhere promising. Wage insurance changes that calculation and gets people back into productive work faster.

Self-employment assistance is another underused tool. AI is genuinely lowering the barriers to starting a business — not just in technology, but in any field where market research, customer communication, content creation, and financial modelling can be assisted by intelligent tools. Supporting workers who want to become founders or sole traders during the risky early months is not charity; it is a cost-effective alternative to prolonged unemployment and the social costs that come with it.

Finally, we need to change the incentive structures facing employers. Right now, announcing large-scale layoffs frequently produces a short-term rise in a company's share price. The market rewards the easy decision. A policy environment that made retraining workers genuinely more attractive than replacing them — through tax credits, through redeployment incentives, through stronger obligations attached to AI productivity gains — would change corporate behaviour more effectively than any number of public appeals to corporate responsibility.

What History Actually Teaches Us About Getting This Right

Optimism about AI and jobs is not irrational. Every major technological transition in history — the industrial revolution, electrification, computerisation — ultimately produced more jobs than it destroyed, raised living standards, and created industries that nobody had imagined beforehand. There is no strong reason to believe AI will be fundamentally different over the long run.

But history also teaches that the transition period matters enormously. The communities and workers who fall through the cracks during a technological shift do not simply wait patiently for the new equilibrium to arrive. They suffer real losses, develop justified grievances, and often make political choices — understandably — that reflect their experience of being abandoned. The political turbulence that has defined much of the Western world over the past decade is not unconnected to the poorly managed deindustrialisation of the 1980s and 1990s.

The difference this time is that we can see it coming. We have the research, the historical precedent, and — if the will is there — the resources to build something better. Post-World War II investment in research and education seeded decades of technological leadership. The COVID-19 pandemic, for all its horror, dramatically accelerated the development and adoption of mRNA vaccines and telemedicine. When the stakes are high enough and the urgency is clear enough, coordinated action at scale is possible.

The question is not whether AI will transform work. It will. The question is whether that transformation is managed in a way that distributes the gains broadly or concentrates them narrowly — and whether the workers displaced in the process have a genuine bridge to what comes next.

A Practical Conclusion: What Needs to Happen Now

The transition to an AI economy is not a problem that can be solved by any single actor. It requires a new compact between government, industry, and educational institutions — one built on outcomes rather than inputs, on flexibility rather than rigidity, and on urgency rather than the comfortable assumption that markets will sort it out.

For governments, the immediate priorities are restructuring funding incentives for training programmes, modernising the unemployment insurance system, and creating tax mechanisms that reward companies for retraining rather than replacing workers.

For businesses, the imperative is to engage honestly and early — not to wait until layoffs are announced, but to work with educational partners and government agencies months or years in advance to build the pipelines that workers will need. Companies that do this will find it serves their interests: a labour market full of newly skilled workers is better for everyone than one full of embittered, economically excluded former employees.

For individuals, the shift in mindset required is significant but not unprecedented. Continuous learning is not a burden unique to our era — it has always characterised successful careers. What needs to change is the infrastructure that supports it, so that learning throughout a career is as accessible, affordable, and socially valued as learning at the beginning of one.

We are not choosing between AI and prosperity. We are choosing between a planned transition and an unplanned one. The costs of the latter — in human suffering, political instability, and ultimately in the regulatory backlash that would slow the technology itself — are high enough to make the investment in planning not just humane, but strategically essential.

Frequently Asked Questions

How many jobs is AI actually likely to replace in the near term?

Estimates vary significantly depending on methodology, but studies from institutions including McKinsey, Goldman Sachs, and the OECD suggest that between 25 and 50 per cent of current job tasks in developed economies are susceptible to automation by AI tools available or in development today. That does not mean an equivalent proportion of jobs will disappear — tasks change faster than whole roles — but it does mean that the skills required for most jobs will shift substantially, often within the span of a single career.

Isn't it true that technology always creates more jobs than it destroys?

Historically, yes — but with important caveats. New technologies do generate new industries and new roles, often on a scale that exceeds what they eliminate. The problem is timing: destruction tends to happen faster than creation, and the workers displaced are often not the same people who fill the new roles. Managing the transition period — the gap between disruption and new equilibrium — is where policy makes the difference between broadly shared prosperity and concentrated suffering.

What is wage insurance and does it actually work?

Wage insurance, or wage top-up, is a policy mechanism that pays displaced workers a portion of the difference between their previous salary and a lower salary in a new job or sector. The idea is to remove the financial penalty for taking a step down in pay to enter a promising new field, accelerating reemployment and career pivots. The United States has operated small-scale wage insurance programmes for trade-displaced workers under the Trade Adjustment Assistance programme, and evidence from those programmes and similar ones in Germany and Denmark suggests they do increase take-up of new employment and reduce long-term unemployment duration.

What can individual workers do right now to prepare for AI disruption?

The most durable strategy is to invest in skills that complement AI rather than compete with it: complex judgment, interpersonal communication, creative synthesis, ethical reasoning, and the ability to supervise and interrogate AI outputs. Practically, this means seeking out employer-led training opportunities, exploring short-form credentials in adjacent fields, and being willing to engage with AI tools directly — understanding what they can and cannot do is itself a marketable skill. Workers who treat AI as something that is happening to them will be more vulnerable than those who treat it as a tool they are learning to use.