For most of the conversation about artificial intelligence, we seem to be asking some version of the same question:
Which jobs will AI replace?
It is an understandable question. It may also be too small.
Jobs are not fundamental laws of economics. Neither are corporations, departments, forty-hour work weeks, career ladders or the idea that most adults should sell their labour to one organization in return for a monthly salary.
They are ways we have learned to organize economic activity. And they can change.
So perhaps the more interesting question is not what AI does to our jobs.
What does AI do to the economic structures that made those jobs necessary in the first place?
To think about where we might be going, it is worth looking at where we came from.
Before the factory
Before the Industrial Revolution, much of economic life in Europe was organized around agriculture, households, small workshops and local trades.
A family might farm while also spinning wool, weaving cloth, making shoes or producing other goods. Craftspeople worked from homes and workshops. In the textile industry, merchants increasingly supplied raw materials to households, collected the finished goods and sold them into wider markets.
This became known as the domestic system, the putting-out system, or, more familiarly, cottage industry.
The popular image can become a little too picturesque: an independent craftsperson happily producing beautiful things beside a roaring fireplace.
Reality was less sentimental. Cottage workers could be dependent on merchants for materials and access to markets, incomes could be precarious, and power was certainly not distributed equally.
But one characteristic matters for our story:
Production was distributed.
The workplace, the household and the means of production were often closely connected.
Even before the steam-powered factory age, Britain had developed extensive networks of manufacturing and home-based production. Historical reconstruction from the University of Cambridge suggests Britain's movement away from a predominantly agricultural workforce toward manufacturing began much earlier than the traditional Industrial Revolution narrative implies.1
Then something changed.
The machine moved production
The Industrial Revolution did not simply give workers better tools. It changed the economics of where production happened.
A spinning wheel could live in a cottage. A large mechanized textile mill could not.
Industrial machinery required capital, buildings, energy, maintenance and coordination. Production increasingly moved toward places where those things could be assembled efficiently.
The worker therefore moved toward the machine.
Factories did more than increase output. They concentrated productive capability.
And industrialization helped accelerate a larger transformation that would eventually give us the modern corporation: large numbers of people, capital and specialized capabilities coordinated inside a single organization.
Today the machine may be a data centre rather than a steam engine, and the worker may sit behind a laptop rather than a loom, but much of the organizational logic remains recognizable.
Consider a modern company.
It may need finance, legal, human resources, technology, marketing, sales, project management, analytics, design, software development, administration and management simply to operate.
We assemble all these specialists inside organizations because coordinating them provides economic advantages.
In 1937, economist Ronald Coase famously asked why firms exist at all. If markets can connect buyers and sellers, why don't individuals simply contract with one another for everything they need?
His answer centred on what economists now call transaction costs.
Finding suppliers costs something. Negotiating agreements costs something. Coordinating work costs something. Monitoring performance costs something. Sharing information costs something.
Sometimes it is simply cheaper and easier to bring all of those activities inside a firm. The optimal size of the firm therefore depends partly on the relative cost of coordinating activity internally versus obtaining it through the market.2
That becomes very interesting when a technology appears that can dramatically reduce the cost of coordination and expertise.
Which brings us to AI.
What if the machine comes back to the cottage?
For the last several decades, powerful digital tools have become progressively more accessible to individuals.
A laptop today gives one person capabilities that once required entire departments.
AI may accelerate that trend by orders of magnitude.
Imagine a capable individual in 2036 with access to AI systems that can perform substantial portions of:
- software development
- research
- accounting
- design
- marketing
- analytics
- administration
- customer support
- contract analysis
- project coordination
- procurement
- sales operations
This is no longer entirely theoretical.
Experimental evidence reviewed by the OECD already finds meaningful productivity improvements from generative AI across tasks including software development, customer support, writing and consulting. The more interesting long-term question is not whether these tools improve individual productivity, but what happens when many of these capabilities become integrated and increasingly autonomous.3
The International Labour Organization's latest work also points toward transformation of work rather than simple wholesale automation as the more useful framing. Exposure is significant, particularly in clerical and highly digitized professional work, but jobs contain mixtures of tasks rather than being indivisible objects that simply vanish.4
So consider a different possibility.
Perhaps AI does not merely allow a company of 5,000 employees to become a company of 3,000.
Perhaps it allows some companies that once required 500 people to operate with 50.
And perhaps businesses that once required 50 people can operate with five.
Eventually, in some fields, perhaps one.
That would change more than employment.
It would change the economically viable size of the firm.
In that world, AI becomes a strange counterforce to industrialization.
The Industrial Revolution concentrated productive capability because individuals could not economically own or coordinate the machinery required for industrial-scale output.
AI could distribute certain forms of productive capability because the machinery of knowledge work becomes accessible from a laptop.
The cottage does not literally return. But some of its economic logic might.
Individuals and very small teams could once again become viable units of production, this time equipped with digital workers, global distribution and capabilities that previously required entire organizations.
We might call this emerging model the neo-cottage economy.
And if it becomes widespread enough to reshape how economic production is organized, we may be looking at a Neo-Cottage Revolution.
There is already a faint outline
We should be careful not to confuse a hypothesis about 2036 with a description of 2026.
Most small businesses using generative AI today are still using it for relatively peripheral tasks rather than fundamentally rebuilding their operating models. An OECD survey published in 2025 found GenAI use among surveyed SMEs was meaningful, but only a minority of those users reported applying it to their core business activities.5
But adoption is moving quickly. Across OECD countries with available data, reported firm-level use of AI rose from 8.7% in 2023 to 20.2% in 2025.6
There are also early signs that AI's employment effects may not be evenly distributed.
Research from Stanford's Digital Economy Lab using payroll data has found employment declines concentrated among younger workers in some occupations where AI is particularly capable of automating tasks, while more experienced workers and workers in occupations where AI acts more as an augmentation tool have fared differently. The researchers are careful about causality and alternative explanations, but it is precisely the sort of structural signal worth watching.7
None of this proves that a neo-cottage economy will emerge.
But it gives us a testable idea.
If H1 is correct, the significant economic story of AI may not simply be fewer people working.
It could be more organizations with fewer people in each.
And that would be a very different future.
But there is a problem
Unfortunately, our neat little theory begins to break almost immediately.
Suppose that in 2036 I have access to extraordinarily capable AI.
I can research your business, build software, analyze financial information, create marketing campaigns, prepare contracts and coordinate delivery with an army of digital agents.
Wonderful.
I announce:
“I am an AI-powered one-person consulting company.”
And my prospective customer replies:
“So am I.”
We have reached the uncomfortable question at the centre of the neo-cottage hypothesis.
Why does the customer need me?
If powerful AI becomes widely available, then access to AI cannot itself remain a competitive advantage.
If my AI can create your marketing campaign, presumably yours can too.
If my AI can analyze your accounts, yours can do that.
If my AI can build software, why hire my software company instead of asking your own AI to build it?
The logic that allowed the individual to compete with the corporation may also allow the customer to bypass the individual.
And suddenly the disruption is larger again.
AI would not simply challenge jobs. It would challenge some of the economic transactions that those jobs and companies exist to perform.
So what are we actually selling?
This is where the HumainX question begins to emerge.
If intelligence becomes abundant, economic value must migrate toward things that remain scarce.
Which means our neo-cottage hypothesis needs refinement.
The future independent producer may not sell AI capability.
AI capability could become table stakes.
They will need to own, know, represent, decide, create, risk or deliver something that the customer's AI cannot simply reproduce.
And that takes us somewhere much more interesting.
The cottage has another possible owner
There is also a darker interpretation of our historical analogy.
The original cottage worker was not always an independent entrepreneur.
In the putting-out system, merchants could own the materials, control access to markets and determine much of the commercial relationship while households supplied the labour.8
The modern equivalent is not difficult to imagine.
I may appear to be an independent AI-enabled business in 2036.
But what if I rent my intelligence from one platform, my computing capacity from another, sell through a third, receive payments through a fourth and depend on algorithms I neither own nor control to reach customers?
I own the cottage.
Someone else owns the loom, the road, the marketplace and perhaps the customer relationship.
That is not economic decentralization.
It is centralized infrastructure wrapped around decentralized labour.
So a neo-cottage economy has at least two possible forms.
Both could happen simultaneously.
The question is no longer who has the smartest machine
The Industrial Revolution reorganized economies around a new source of productive power.
AI may be doing the same.
But this time the scarce resource being transformed is not primarily mechanical force.
It is intellectual capability.
If that capability becomes radically cheaper and more abundant over the coming decade, we should expect consequences far beyond the automation of individual jobs.
We should expect pressure on the boundaries of companies.
On the value of expertise.
On the relationship between employee and employer.
On the balance between labour and ownership.
And ultimately on the question of what one person can economically offer another.
That is the experiment HumainX will follow toward 2036.
We will start with a hypothesis:
But we have already discovered its problem.
If everyone has access to the same extraordinary productive intelligence, being an AI-enabled individual may not be very special at all.
So before we predict a world of million-dollar one-person companies, we have another question to answer.
Perhaps the more important one.
Ten years from now, will most people still earn the majority of their income from a single employer?
- Yes, largely unchanged
- Yes, but companies will employ far fewer people
- No, multiple income sources will become the norm
- No, traditional employment itself will become much less important
- I have absolutely no idea
Stay with the question.
New arguments, counterarguments, reader predictions and evidence as we work toward understanding the economy of 2036.
Sources & further reading
- University of Cambridge — A nation of makers: industrial Britain began earlier than we thought
- Ronald Coase — The Nature of the Firm (1937)
- OECD — Unlocking productivity with generative AI: evidence from experimental studies
- ILO — Generative AI and Jobs: A Refined Global Index of Occupational Exposure
- OECD — Generative AI and the SME Workforce
- OECD — AI adoption by firms continues to expand
- Stanford Digital Economy Lab — Canaries in the Coal Mine?
- Corporate Finance Institute — Cottage Industry
HumainX is an ongoing Mindsgate exploration of work, ownership and economic life when intelligence is no longer scarce.