Dr Roman Yampolskiy, an AI safety researcher, appeared on The Diary of a CEO recently and made a prediction that has been circulating widely since. By 2030, he argues, AI combined with humanoid robotics will make it "make no sense to hire humans for most jobs." He puts potential unemployment at levels approaching 99%, not the 10% that would already be frightening.
Even allowing for the fact that academic predictions about AI timelines have been consistently wrong for thirty years, the direction of travel is clear enough. BCG's 2026 workforce study found that 49% of white-collar tasks can already be performed by current-generation AI tools. McKinsey estimates that 30% of work hours could be automated by 2030. These are not fringe views.
Dr Yampolskiy identifies five categories of work that might survive. They are worth looking at, because they tell you something about the shape of the future economy, and about where the real opportunity sits.
The five categories
1. The human touch, for those who can afford it. Yampolskiy suggests that certain professional relationships will retain a human preference. He uses Warren Buffett as an example. Buffett would not switch to an AI accountant, even if the AI was faster and more accurate. People "like traditional ways of doing things." The implication is that this is a luxury market. The human accountant survives not because they are better, but because enough wealthy clients prefer a person.
2. A fetish for human-made products. His word, not mine. A small subset of consumers will pay more for handmade, locally sourced, artisanal goods, in the same way that some people currently pay more for handcrafted furniture than flat-pack equivalents. Yampolskiy is honest about the scale. This is "almost a fetish" and will not support large numbers of jobs.
3. Jobs requiring lived human experience. Counselling, therapy, and certain forms of care work. People who have experienced grief, addiction, recovery, or parenthood bring something that an algorithm cannot replicate, however sophisticated the model. This is the strongest category in the list, because it is grounded in something fundamental about what humans value from other humans.
4. AI regulation and oversight. Someone will need to supervise AI systems, at least during a transition period. Yampolskiy is sceptical that long-term containment is feasible, but believes that human oversight now will slow the rollout from five years to fifty. This is essentially a new industry built around managing the technology that is displacing everything else.
5. AI intermediaries. People who understand both how AI works and how to explain it to non-technical humans. If a business is embedding AI into its operations, it needs translators. Not engineers, but interpreters. People who sit between the technology and the people who need to use it.
What the list actually tells you
Read the list again and notice what all five categories have in common. Every single one is either a luxury (human touch, artisanal goods), a niche (therapy, AI oversight), or a transitional role (AI intermediary) that exists only because of the disruption itself.
None of them are the mass-employment categories that currently define the UK economy. There is no manufacturing, no logistics, no retail, no hospitality, no financial services, no administration, no construction. The five "safe" categories collectively employ perhaps 3% to 5% of the UK workforce today.
The unstated conclusion is that the question "which jobs will survive AI?" might be the wrong question entirely.
The honest objection
Before going further, there is an obvious counter-argument that needs addressing. If AI displaces the workforce, who buys the products and services that businesses produce? A company that sells software to accountants cannot survive if all the accountants are unemployed. A plumbing supply business cannot thrive if plumbers have been automated out of existence.
This is a real concern, and anyone suggesting that business ownership is a simple immune shield against AI disruption is being dishonest. Macro-level economic disruption affects everyone. Business owners are not exempt from the cascade effects of mass unemployment.
But here is the distinction that matters. Business ownership does not make you immune to disruption. It gives you more levers to adapt to it.
An employee whose role is automated has one outcome: income loss. They can retrain, but they are retraining for a different position in the same disrupted system. Their response is reactive, and their timeline is set by someone else's decision to replace them.
A business owner whose market is disrupted can restructure the entire enterprise. They can pivot to a different customer base. They can automate their own processes and reduce costs. They can acquire complementary businesses at distressed valuations. They can change their product, their pricing, their service model, and their go-to-market strategy. None of these are guarantees of survival, but they are options. The employee has none of them.
The businesses most likely to survive AI disruption share certain characteristics. They serve other businesses rather than consumers, because B2B demand is more resilient than consumer spending during transitions. They own physical assets or infrastructure that cannot be replicated digitally. They have recurring contracts or relationships that create switching friction. And critically, they can integrate AI into their own operations to reduce costs faster than their revenue declines.
Owning a business does not protect you from the macro economy. But it gives you agency within it. Employment gives you none.
Why ownership is structurally different in an AI context
There is a second advantage to ownership that is specific to AI disruption, and it is worth stating clearly.
When AI automates a task, the value of that task drops to near zero. If you are an employee whose job is that task, your income drops to near zero. If you are a business owner whose process includes that task, your cost drops to near zero and your margin improves. The same technology that destroys employment income creates operational leverage for owners.
This is not a theoretical point. Companies across the UK are already using AI to handle bookkeeping, customer service, scheduling, content production, and data analysis. The employees who did those tasks are being made redundant. The business owners who deployed the tools are seeing their operating costs fall.
The implication is uncomfortable but clear. In an economy where AI displaces labour, the returns shift from the people who do the work to the people who own the systems that do the work. This is not a moral argument. It is a structural observation about how technology transitions distribute value.
The ownership path in practice
There are approximately 808,000 UK businesses with a sole director aged 60 or above. Many are profitable, established, and have no succession plan. These are companies with real customers, real revenue, and real operational history, sitting with owners who will need to exit within the next five to ten years.
Buying one of these businesses is not a hedge against AI. It is a direct investment in cash flow and operational control that exists outside the employment market. Whether AI displaces 10% of jobs or 50%, businesses still need to be run. Customers still need service. Operations still need management. The difference is that the owner can deploy technology to improve margins, while the employee can only hope their role is not the one that gets automated.
The constraint has historically been finding the right business. That is no longer the barrier it was. The infrastructure to identify, enrich, and contact business owners at scale now exists. The barrier is the decision to act.
The real takeaway from the five jobs list
The five jobs list is interesting as a data point. It tells you something about where human preference still matters. But the deeper insight is not about which jobs survive. It is about the difference between being inside a system that is being disrupted and holding a position that lets you adapt within it.
If you are worried about AI and your career, retraining is a reasonable response. But it is worth asking a bigger question. Instead of training for one of five surviving job categories, competing with everyone else who had the same idea, would you rather own the company that employs people in them?
The answer is not obvious. Ownership carries risk, responsibility, and no guarantees. But in a world where the alternative is hoping your specific role is one of the five that survive, it is a question worth taking seriously.