About 3.5 billion people.
That is the approximate size of the global workforce today.
Some are specialists, doctors, lawyers, scientists, engineers. Some do manual work, construction, farming, manufacturing, and similar. Many are in the service economy, providing support, administration, analysis, coordination, and oversight.
This workforce exists within a much larger pool. Roughly 4 to 6 billion people worldwide are of working age, a number that grew from about 4 billion in the early 2000s and is projected to approach 6 billion by the 2040s, peaking later this century. Not all of them are employed, but all of them depend on the same system, paid work as the primary mechanism for distributing income.
That system is now changing rapidly, propelled by the advancement of GenAI efforts.
Some roles have already been or are being actively replaced by AI. On one hand, it seems like another technology advance, no different than phone operators who got replaced by automated systems decades ago. Customer support, basic analysis, content generation, QA, monitoring, scheduling, and first-pass decision making are all examples of roles being replaced by AI daily. It’s in the headlines, and companies don’t hide it, they brag about it instead. In some cases, people are not being laid off. They are simply not being replaced, as they retire or leave for another opportunity. The role disappears, the budget does not. It gets redirected, often into more AI investments. After all, data centers aren’t cheap. Nvidia chips aren’t cheap. Even Google’s new less expensive chips will certainly not be a true discount bin acquisition. Oracle, for instance, is already scaling back new data center projects, combined with their stock pulling back. Just to think that recently Larry Ellison became richer than Elon Musk, even if only momentarily. The hype is real, but so are the expenses. If recent changes are to be followed, it’s safe to presume that layoffs may come next.
Other, more nuanced and complex roles, are still waiting, not because they are safe, but because the hardware is not yet cheap, reliable, or widespread enough. Jobs, like construction, logistics, physical maintenance, as well as people-facing ones, like caregiving, waiting on tables, concession stand attendants and so on are not ready to be replaced by Optimus robots just yet, although Miso Robotics automation is already replacing some cooks. Overall, for those positions, the delay is technological for now, not philosophical.
This trend is often framed as ‘the end of employment as we know it.’ That framing is somewhat correct, but it also misses much of what is actually happening.
What companies are doing is rather predictable and not at all new to the economy. They are stopping the use of people for work machines can already do and reinvesting the savings into tools that can do even more of it next quarter.
The issue doesn’t become ‘AI replacing people’, but rather ‘old roles being eliminated, while new roles aren’t being created’.
The causes of the latter can easily be traced to the last 5 or so years, starting with the 2020 Covid pandemic. The influx of newly printed money, meant to offset the devastating effects of the once-in-a-century event created an unimaginable number of new millionaires, produced unreal dividend payments, and increased often unjustified valuations, among other factors. It also drastically increased the national debt, with interest payments alone on it totaling about $1 trillion annually at this time. Once the financial subsidies ended, people weren’t ready to go back to making less money and being more realistic. One clear side effect – inflation made it often unsustainable to even survive on pre-pandemic income, while real estate, food, cars, energy, and just about everything else largely doubled in price in many cases, if not seen more drastic price increases. That alone makes new role creation too unaffordable for most employers. The choice between letting the stock price drop due to skipped stockholders’ awards or lack of AI investment and procrastinating on the ‘issues of tomorrow’ seemingly has been made. Companies are choosing profits and hype, not sustainable living, very much in line with the ballooning national debt to accommodate tax breaks for the rich and social services for the poor.
Some companies are trying the speedy version of reskilling for many. We are already seeing this logic applied at scale. Accenture announced in early December 2025 a broad, multi-faceted partnership with OpenAI, equipping tens of thousands of its professionals with ChatGPT Enterprise and launching a joint AI client program. The message is clear. Reduce friction, increase leverage, push more work through fewer people, supported by increasingly capable tools, and try to retain experienced professionals by equipping them with new tools and skills.
It is critical to see a major wrinkle in available solutions. While small employers are suffering through this stage, largely unable to afford the trendy changes, major firms, from BigTech to major brokerage, are doing just fine. Their earnings have risen far faster than even expenses, their stocks are having the wildest of times on Wall Street, and their executives have never seen current levels of compensation. That creates little incentive for them to change things, at least till market conditions change. Employees are suffering, being overworked, and laid off staff often take 6+ months to find even ordinary roles, while the rate of 2025 layoffs hasn’t been this high since the beginning of the pandemic, but they do not get to affect this game. Only employers have any power here, and they are choosing tools that outperform biology in digital environments. No fatigue. No context switching. No memory decay. Just iteration, parallel execution, and scale. A recent Stanford experiment makes this shift harder to ignore. An autonomous AI agent was allowed to operate inside a controlled university network for roughly sixteen hours, performing reconnaissance, vulnerability discovery, and exploitation tasks. It outperformed most human cybersecurity professionals doing comparable work, at a fraction of the cost. There was no genius moment, no creativity spike. The advantage came from operating entirely inside a digital environment at machine speed. No fatigue. No context switching. No ego. Just continuous iteration, parallel execution, and perfect recall. When technology is allowed to play technology’s game, the comparison stops being fair. Cybersecurity is written by humans, but it runs at machine tempo. Once agents are allowed to live fully inside that world, biology becomes the bottleneck.
The real question is not whether AI can replace parts of human work. That answer has never been this obvious and easy to spot, even compared to the emergence of personal computers or the internet.
The real question is whether companies actively create and protect human roles that are not economically rational to automate, and whether they do so fast enough. Not passively. Not eventually. Intentionally. This isn’t about social norms or the greater good. It’s demand-based, as it has always been in a capitalistic society. The critical question becomes whether there is enough demand for human-only or human-centered roles for billions of working-age people to transition into within the next five to ten years. It may seem like a short period of time to see major transformation, but it becomes more realistic, when viewed through the prism of changes from the last three years alone, since the first version of ChatGPT upended workforce needs, and considering the over a million layoffs in 2025 alone, with the year still not at its end. Even if it may seem that companies don’t care about people and only focus on profits, they still need consumers to make money. Consumers still need income to consume. It is reasonable to mention that the current economy is holding up largely due to the top 10% of earners’ spending, even as the middle and lower class have pulled back spending drastically in recent times.
Efficiency that removes labor faster than new human roles are created does not lead to a new equilibrium. It erodes the demand base of the system itself. This is not an argument for communism. Nobody is moving there. It is not a moral critique of technology.
It is a question of economic design. Either companies begin designing roles, incentives, and responsibilities where humans remain essential, or they continue optimizing in ways that quietly undermine their own market. In theory, companies will stop the automation/efficiency trend before it begins to hurt their bottom line, but in reality, it’s no different from Microsoft AI boss promising development halt if AI tech poses existential risk. This is a good place to remember that even CEOs no longer control companies. Investors and shareholders have virtually unlimited say, whether through Board of Directors decisions, private equity and venture funds demands of return on investments, or even activist shareholders, looking to increase their earnings.
With all factors in mind, how will private and public entities satisfy the unprecedented hunger for earnings; keep AI advancements in mind, even if per Yann LeCun, it may come through ‘world models’, not LLMs; reskill current employees; create new human-level responsibilities and roles, while keeping them away from the reach of AI for no less than one to two decades; and not crumble, when the next recession hits and the government, already overburdened by national debt, is unable to save them?
Those are the real challenges of today. Nobody is willing, or able, to give up the benefits of the system, and nobody is looking to stagnate either. As Lewis Carroll said in his Alice in Wonderland saga long ago, ‘…it takes all the running you can do, to keep in the same place. If you want to get somewhere else, you must run at least twice as fast as that!’. The system needs to fuel itself, while looking into the future, and the present guardrails are borderline nonexistent. Just look at the new executive order from President Trump, demanding unified AI compliance among all states, while limiting state regulations of the AI development progress.
It’s important to note that an undesirable outcome is not inevitable, and if it arrives, it won’t be accidental either. Everything, from seemingly unpredictable to planned influencing factors is the result of incentives, timelines, and choices being made right now. The question is not whether AI will advance, as it’s moving forward at incredible speeds, and the already made investments may not bring a dollar for dollar return but are certain to transform the economic and workplace landscape. The question is whether companies intentionally design human-essential roles before demand erodes beyond repair. That is the trade-off most are avoiding even in consideration and certainly in action.
