A recent episode of The Economist’s Checks and Balance on how America should prepare workers for AI sent me down a rabbit hole. The discussion went back through earlier periods of workplace upheaval and the institutions the U.S. built to help people adjust. Looking at that history, there is a fairly consistent pattern: when the labor market changed dramatically, American policymakers tried to figure out where demand for workers was moving and then help people get there.
At the end of World War II, the U.S. Department of Labor feared that as many as 15 million returning servicemen and women could end up unemployed as the war economy wound down. The catastrophe never arrived at anything close to the scale feared, but the country still built an enormous transition apparatus around education, training, and reintegration, most famously through the GI Bill. The same instinct appeared again as automation spread through American industry in the 1950s and 60s, and later as globalization displaced workers in import-exposed industries. The programs changed, and none worked perfectly, but the underlying idea was remarkably durable: if the economy needs people to do something different, help them move toward the work that now needs doing.
AI may make that transition look a little different.
I don’t think AI makes humans unnecessary at work. In The Crossroads, I argued that the final layer of many jobs remains stubbornly human: the judgment call, the customer relationship, the difficult feedback conversation, the decision somebody ultimately has to own. AI can do vastly more of the research, analysis, preparation, and execution leading up to those moments, while humans remain accountable for what happens next.
What changes is how many people are required to get there.
A corporate team of thirty might eventually need ten people sitting on top of an enormous amount of machine work. Those ten become dramatically more productive. Some of the others will find another corporate job, some will retrain into healthcare, skilled trades, infrastructure or professions we have not invented yet, and some may simply decide that another corporate role is not what they want next.
AI will create employment of its own too. Data centers need electricians, construction workers, power engineers and technicians; U.S. data-center employment grew from roughly 306,000 people in 2016 to 501,000 in 2023. Frontier AI companies need scientists, engineers, and entirely new kinds of specialists.
But the more interesting question is what happens to someone who is still perfectly capable of creating value, but is no longer needed in quite the same way inside a large organization.
For many more people, I think the answer could be working for themselves.
The word entrepreneurship tends to send the imagination toward venture capital, pitch decks, and somebody trying to build a billion-dollar software company. Most people who work for themselves will probably do something much less dramatic.
A great cook might build a ten-table restaurant around a cuisine they have spent twenty years perfecting. A football obsessive could make a living coaching young players. A former executive with unusually deep knowledge of one industry might consult for six companies instead of working full-time for one. Someone else might run a tiny game studio, build social experiences for communities and brands, restore vintage motorcycles, teach music, or create a small hospitality business around a place they know unusually well.
There are many versions of this future:
Craft and product businesses: food, furniture, fashion, repair, restoration, specialty manufacturing, art and design.
Creative and digital businesses: game development, small media studios, interactive experiences, niche software and digital products.
Relationship and experience businesses: coaching, tutoring, fitness, hospitality, community building, social experience management, events and local services.
Passion-first solo livelihoods: writers, musicians, creators, researchers and specialists combining several sources of income without necessarily wanting to build a larger company.
Some of these are true small businesses with employees, leases and physical operations. Others are much lighter-weight solo practices built around expertise, reputation or a small set of recurring clients. The economics are different, but the underlying idea is similar: the person owns more of the work itself, rather than simply supplying labor into somebody else’s system.
That is also what makes this different from much of the gig economy. A gig worker may technically be independent while the platform still controls the customer, pricing, and distribution. The version I find more interesting gives the individual much more ownership of the economic asset itself: the coach owns the relationship with families, the consultant owns the client base, the game developer owns the product, the restaurant owns its identity.
And “being your own boss” does not mean being free of customers, competition or risk. It means having more control over what you build, who you serve, how large you want to become, and whether you want to change direction altogether.
The obvious pushback to this idea is that people have wanted to be their own bosses forever, and those with the right resources have always been able to try. The interesting question is why this could become possible for many more people now.
The internet solved one of the biggest historical disadvantages of being small: finding enough customers.
Thirty years ago, somebody teaching a niche subject largely needed enough interested people within driving distance. A craftsperson depended heavily on local demand. A specialist consultant needed the right companies to somehow know they existed.
Today, a teacher can find 200 students scattered across fifteen countries. A tiny brand can find customers nationally. A guide offering an incredibly specific kind of trip can be discovered on another continent. A consultant can find the twelve companies in the world for whom their weirdly narrow expertise is enormously valuable.
The internet made the addressable market for being unusual much larger.
Software already gave tiny businesses access to payments, accounting, CRM, ecommerce, scheduling and analytics that once required much more infrastructure. But the owner still had to operate all of it. SaaS gave small firms the tools of a large company, not the people.
AI starts closing that gap. A restaurant can get help with reviews, promotions, supplier analysis and customer communication. A coach can have scheduling, CRM and admin support. A consultant can have a research analyst. A tiny game studio can get help with community management, localization, support and marketing.
Some of these functions can already be handled continuously by persistent agents, with the owner stepping in for judgment, taste, relationships and important decisions.
That is the bigger economic shift: the minimum efficient size of the firm gets smaller.
Not everywhere. Capital-heavy industries like airlines, hospitals or semiconductor fabs will still need scale. But in services, education, creative work, game development, hospitality, niche commerce and many small businesses, far less organization may be needed around the person doing the valuable work.
Large companies will remain extraordinarily good at things where scale is genuinely useful. Semiconductors, cloud computing, logistics, energy, pharmaceuticals, commodity manufacturing and plenty of other industries probably become even more concentrated and efficient.
I don’t think anyone is looking for artisanal electricity.
But there are many categories where customers value exactly the opposite. Restaurants. Education. Fitness. Travel. Fashion. Entertainment. Hospitality. Design. Coaching. Advisory work. I may willingly pay more for that chef’s food, that coach’s attention, or that designer’s taste.
Historically, boutiques could command that premium while also suffering from terrible economics. Personalization was expensive because all the boring machinery surrounding it was expensive too. AI can reduce the marginal cost of marketing, administration, support, analysis, and coordination without removing the human specificity the customer is paying for.
Higher willingness to pay. Lower cost to serve. Suddenly, a much larger universe of weird little businesses clears the threshold for viability.
The final part of the “why now” argument is more speculative and hopeful, but perhaps most important.
Working for yourself becomes much easier if staying alive becomes cheaper.
Imagine basic healthcare in an AI-rich economy. Routine medical intelligence could become ubiquitous: a system that knows your history, monitors your health, answers basic questions instantly, triages symptoms, and brings in a doctor precisely when scarce human expertise is required. AI could reduce administrative overhead throughout healthcare and accelerate drug development. Great doctors and complex procedures will remain scarce and expensive, but a huge amount of basic medical intelligence does not have to be.
Food could move through its own productivity curve as autonomous equipment, precision agriculture, better crop prediction, robotic warehouses and increasingly efficient logistics reduce the cost of producing basic calories. Energy could become dramatically cheaper through better engineering, grid management, storage, new materials and whatever generation technologies the next two decades produce. Cheap energy then lowers costs everywhere else: compute, transportation, food, manufacturing and construction.
The future could therefore contain a strange combination. A townhouse overlooking Central Park remains insanely expensive. A meal from a famous chef remains expensive. The attention of the best surgeon in the world remains scarce.
But basic calories, basic medical intelligence, energy, manufactured goods and a functional place to live could all become much cheaper. This, however, is a leap of faith from a techno-optimist.
If somebody needs substantially less income to maintain a decent life, experimenting with independent work becomes far less reckless.
Which brings us to the part technology alone will not solve.
The historical transitions I started with were not left entirely to the labor market. Governments played an active role in making them possible: funding education and retraining, providing transition support, building institutions around new skills, and reducing the cost of moving from one kind of work to another. The GI Bill is perhaps the clearest example. When the destination of work changed, public policy helped build the bridge.
If working for yourself becomes a more important destination in the AI economy, governments may need to build a different kind of bridge. Instead of focusing almost exclusively on retraining someone for their next employer, workforce policy could also make it easier for them to create their own source of income: lowering the risk of leaving payroll, making small amounts of capital and shared infrastructure accessible, simplifying the bureaucracy around starting something, and teaching the commercial skills required to turn expertise into a viable business.
The first requirement is a floor. Healthcare and basic benefits should become less tightly connected to employment, while temporary transition support can give someone enough runway to test whether an idea has real demand.
This is not about making every passion economically viable. If nobody wants the table, eventually the woodworker should stop making tables. The objective is simply to make one failed attempt less likely to wreck the family’s finances.
Most people starting these businesses do not need venture capital. They may need $15,000 of equipment, access to a commercial kitchen twice a week, a field to coach on, or a studio they can rent by the hour.
Microloans, shared facilities, simpler licensing and practical education can turn large upfront commitments into smaller variable costs. Workforce programs could teach pricing, sales, cash flow, negotiation and AI fluency alongside the traditional focus on qualifying people for another job.
If millions more people work independently, markets will build around them.
AI systems can bundle bookkeeping, marketing, scheduling, analytics and customer support. Fractional accountants, lawyers and designers can serve dozens of tiny firms. Logistics networks, group purchasing, marketplaces, lending and insurance can give small operators capabilities that once required much more scale.
Instead of rebuilding a corporation around themselves, people can increasingly rent the pieces of one they need.
If AI allows large companies to generate far more output with fewer people, while more individuals are trying to establish small firms and solo livelihoods, the tax system should reflect that too.
To me, that is a much more attractive future than one where AI creates enormous corporate abundance and the primary answer is simply to distribute enough welfare or UBI for everyone else to consume it. I would rather preserve the basic engine of capitalism: people becoming excellent at something, competing, building businesses and capturing real upside when they create something others value.
It might mean making the contest easier to enter. Small independent earners should face simpler compliance and lighter tax burdens while they establish themselves, while a greater share of common costs can come from the larger pools of profit and capital income that automation creates.
Put all of this together, and the future looks different from either of the two versions we usually hear about - AI does not leave employment untouched. Nor does it make human economic participation obsolete.
Large companies become much more productive. They employ fewer people per dollar of output, and those people sit on top of increasingly capable machine labor. The things that genuinely benefit from scale become extraordinarily cheap and abundant.
Meanwhile, being small becomes much easier.
Some people build conventional small businesses and eventually hire employees. Others construct passion-first solo livelihoods: teaching three days a week, consulting occasionally, running a paid community, performing on weekends, making something specialized, or combining several of those things into a good living. The boundaries between “job,” “business,” “gig,” and “profession” become fuzzier because the administrative burden of coordinating all of it becomes tiny.
And none of this requires abandoning the basic economic bargain underneath capitalism. People still have to make things other people want. They still compete. They still fail. Somebody who becomes unusually good at something can earn more because other people genuinely prefer what they produce.
At its simplest, an economy is people doing useful things for one another. I create something you value, earn money for it, and use that money to buy something somebody else has become good at creating. That exchange gives people a reason to improve, specialize, experiment, and occasionally become exceptional.
AI does not have to weaken that system. It may allow far more people to participate in it on their own terms.
P.S. If (when) robotics enters the chat and starts collapsing the physical cost side of this equation too, I will almost certainly be back with a sequel.
If you liked the content and would like to support my efforts, here are some ways:
Subscribe! It’ll help me get new updates to you easily
Please spread the word - it’s always super encouraging when more folks engage with your project. Any S/O would be much appreciated.
Drop your constructive thoughts in the comments below - point out mistakes, suggest alternate angles, request future topics, or post other helpful resources
Thanks for reading The Lay of the Land! Subscribe for free to receive new posts and support my work.