Admissions from inside Meta expose the limits of rapid workforce replacement, the White House views artificial intelligence as a contest for global influence with China, and Bill Gates warns of social disruption that could reshape work, wealth and security.

The question surrounding artificial intelligence is no longer whether it will change the world, but what kind of world it will produce, who will bear the cost of the transition and who will capture its benefits.

Inside technology companies, a gap is emerging between promises of digital agents capable of taking over human work and the systems’ actual operational performance. In major capitals, governments increasingly regard AI as a pillar of economic and military power. At the societal level, warnings are mounting that the transformation may unfold faster than labour markets and public institutions can adapt.

These developments intersect across three dimensions: an operational dimension in which companies are testing the limits of automation, a geopolitical dimension defined by competition for technological supremacy, and a structural dimension concerning the future of employment, the distribution of wealth and social stability.

Meta and the Limits of Rapid Replacement

Recent developments at Meta offer a revealing example of the distance between the idea of AI agents capable of performing entire jobs and the practical difficulty of turning that vision into a reliable digital workforce.

During an internal meeting in July, CEO Mark Zuckerberg acknowledged that the development of AI agents had not accelerated as executives had expected and that the anticipated benefits of reorganising the company around artificial intelligence had yet to materialise.

His comments came after Meta laid off around 8,000 employees and reassigned approximately 7,000 others to AI-related teams, while projecting spending of up to $145 billion on AI infrastructure this year. Zuckerberg said the company hoped to see clearer returns from those investments over the following months.

The admission does not mean that the AI-agent project has failed or that Meta is retreating from its investments. It shows, however, that moving from a model that answers questions or generates text to one capable of independently, accurately and safely managing an entire sequence of tasks is harder than marketing demonstrations often suggest.

Work inside an organisation is not merely a collection of isolated tasks that can be converted into software instructions. It also relies on contextual understanding, accumulated tacit knowledge, the ability to assess exceptions, coordination between people and accountability when information is incomplete or contradictory.

AI systems may be able to perform large portions of a job while remaining unable to assume the entire role, particularly where human judgement, undocumented knowledge, or legal and ethical responsibility are involved.

This interpretation is consistent with findings by the International Labour Organization, which estimates that one in four jobs worldwide is exposed to generative AI to varying degrees. The organisation nevertheless concludes that the transformation of tasks, rather than the complete disappearance of jobs, remains the more likely outcome at this stage.

The question therefore shifts from “When will AI replace employees?” to a more practical one: how can work be redesigned so that machines handle the tasks they perform well while humans remain responsible for context, judgement, expertise and accountability?

From a Productivity Tool to a Measure of Global Power

While companies test the technology’s limitations in the workplace, the United States, China and other powers increasingly treat AI as part of the infrastructure of global influence.

In its AI Action Plan, the administration of President Donald Trump described the technology as capable of transforming the global economy and altering the international balance of power. It tied continued US economic and military leadership to America’s ability to win the AI race.

The US approach centres on accelerating innovation, expanding data centres, energy supplies and technological infrastructure, promoting American systems abroad, and reducing regulations that the administration considers obstacles to corporate competitiveness.

In March 2026, the White House introduced a national legislative framework calling for the removal of outdated or unnecessary regulatory barriers, the prevention of conflicting state-level rules and expanded workforce training. At the same time, it acknowledged concerns involving children, intellectual property, electricity prices and public trust.

This approach shows that governments do not necessarily begin with the same questions as companies or societies. States are concerned with who controls the most powerful models, advanced semiconductors, data centres, energy supplies and technical standards that other countries may eventually adopt.

From this perspective, slowing down unilaterally may appear more dangerous than continuing at speed, because a country imposing strict limits on its own companies fears giving foreign competitors an opportunity to advance.

Yet the logic of the race creates a clear paradox: the more strategically important technological supremacy becomes, the less willing participants are to slow down or coordinate, even when they recognise that rapid development carries shared risks.

In this respect, AI increasingly resembles an arms race. No participant may consider unrestricted acceleration ideal, but each fears being the first to apply the brakes.

Bill Gates: Disruption Will Not Be a Temporary Economic Crisis

Bill Gates offers a third perspective, extending beyond operational setbacks and calculations of geopolitical dominance to the structure of society itself.

In an essay published in August 2026, Gates described the transition into the AI era as likely to become one of the most turbulent periods in human history. He warned that the world still lacks an adequate plan for managing its consequences.

Gates argues that comparing AI with personal computers or earlier waves of automation may be misleading. People had to learn how to use previous machines and software and reorganise their work around them. AI, by contrast, can operate on devices that already exist, communicate through natural language and learn from the same material used to train human employees.

In other words, people do not have to adapt entirely to the machine this time because the machine can increasingly adapt to human language and working practices. That characteristic may enable AI to spread considerably faster than previous technological transformations.

Gates expects the effects to reach law, customer service, medicine, software and manufacturing, with entry- and mid-level jobs among the most exposed. New roles will emerge, but he warns that they may be fewer in number or require skills that take years to acquire.

The fundamental difference is that AI-driven unemployment may not resemble cyclical unemployment, which recedes when economic growth resumes. If AI systems become permanently capable of performing an employee’s tasks at lower cost, the result will not simply be a temporary downturn but a structural transformation of how economies distribute income, employment and social status.

A Tool for Equality or a Machine for Concentrating Wealth?

Gates does not offer an entirely pessimistic vision. He believes AI can accelerate the discovery of medicines and vaccines, improve healthcare and education, assist farmers and expand access to knowledge and services in poorer communities.

But he places those opportunities alongside corresponding dangers: fraud, deepfakes, disinformation and surveillance, as well as the ability of attackers to discover cybersecurity vulnerabilities and target hospitals, banks, water systems and electricity grids.

He also warns that the same technology that helps scientists develop medicines and vaccines could make it easier to design dangerous biological agents. Tools that empower individuals could simultaneously enhance the ability of governments and armed actors to conduct surveillance and use force.

This leads to Gates’s most consequential proposition: AI could become the greatest equalising force ever invented or the worst new source of injustice. The outcome will be determined not only by the power of algorithms, but also by who owns them and how the resulting gains are distributed.

If productivity gains flow primarily to a small group of companies and investors while millions lose their jobs and incomes, returns on capital will rise as wages weaken, tax revenues decline and demand for public assistance grows.

Gates therefore calls for early consideration of more flexible social safety nets, retraining programmes and reforms to taxation. He has also proposed taxes that could slow the rush to replace human labour while generating resources to support people affected by the transition.

Three Timelines Moving at Different Speeds

Comparing Meta, the White House and Gates reveals three different timelines for artificial intelligence.

Corporate time is measured in months, investment returns and the ability to turn models into reliable products. Government time is measured in years of competition for economic and military power. Social time extends across generations, because the loss of a job, the collapse of a career path or a widening wealth gap cannot be repaired with a new software update.

Within companies, fully replacing people remains harder than expected. Between states, these temporary setbacks are unlikely to halt the race. At the societal level, failing to prepare may prove more dangerous than miscalculating when the technology will arrive.

A Social Contract for the AI Era

The conclusion is neither that AI is merely a bubble nor that it is an autonomous force inevitably destined to eliminate human work. More likely, it is a powerful technology advancing unevenly—excelling at some tasks and struggling with others, yet improving fast enough to force companies, governments and societies to reconsider their assumptions.

The central challenge is to create a new social and technological contract defining who is accountable for decisions made by AI systems, who owns the productivity gains, how people who lose their jobs will be protected, and which boundaries geopolitical competition must not be allowed to cross.

The real choice is not between innovation and fear. It is between a transition managed by societies through clear rules and one imposed by corporate speed and international competition after its social costs have already become unavoidable.