This week, I listened to an interview with Chinese technology investor Wang Yuquan, who made a prediction designed to attract attention: the AI bubble could burst around 2029.
The exact year matters less to me than the reasoning behind it. Wang is not arguing that AI has little value. He is arguing almost the opposite: AI’s long-term potential is so significant that we may be pricing years of social and economic change into the present.
We can already see AI writing, coding, analyzing data, operating software, and carrying out sequences of tasks. From there, it is tempting to jump ahead and assume that entire professions will soon disappear, companies will rapidly reorganize themselves, and everyone will gain abilities that once seemed out of reach.
But technical capability, organizational adoption, and economic value do not arrive at the same time.
A system may be able to perform a task long before a company knows how to build it into a reliable workflow. A company may adopt AI long before its employees understand how to work with it. The technology may spread widely before most people see a meaningful improvement in their income or daily lives.
When we turn “this may happen” into “this will happen,” and “this will happen eventually” into “this will happen very soon,” we are no longer responding only to the technology. We are responding to our own imagination.
That is where a bubble begins: in the distance between what a technology may become and how quickly we expect the future to arrive.
The early Industrial Revolution followed a similar emotional pattern. Mechanization brought greater productivity, but it also created fear, resistance, and confusion. People could see old forms of work disappearing before they could understand the new jobs, industries, and ways of life that would eventually emerge.
Over time, industrial products became part of everyday life, new institutions formed around them, and society gradually adjusted. The technology did not fail; expectations and social structures simply needed time to catch up.
An AI bubble bursting would not necessarily mean that AI had failed either. It could simply mark the moment when exaggerated expectations give way to practical use.
Tasks Change Before Professions Disappear
As I explored in last week’s article , AI is more likely to replace tasks before it replaces entire professions. The first things to change are usually the repetitive, standardized, and clearly defined parts of a job whose results can be checked easily.
A profession is rarely one indivisible skill. A doctor does more than recall medical knowledge, a teacher does more than deliver information, and a programmer does more than write lines of code. Their work combines routine tasks with interpretation, communication, judgment, and responsibility.
As AI takes over some of those tasks, an entire profession does not vanish overnight. Its internal structure changes.
When computers first entered offices, software absorbed some of the typing, formatting, recordkeeping, and information-processing work traditionally performed by secretaries. The role did not immediately disappear, but its emphasis shifted toward coordination, communication, and the handling of more complicated situations.
AI is likely to produce a similar shift on a much larger scale. Work that can be described, repeated, and verified will increasingly move toward machines, while people will be pushed toward defining goals, designing processes, handling exceptions, and evaluating results.
Wang compares large language models to electricity during the Industrial Revolution and AI agents to stations along an assembly line. What matters in this comparison is not whether AI literally resembles electricity, but that a technology begins to transform production only when it becomes part of a continuous process.
Electricity did not complete the Industrial Revolution by lighting a single lamp. Its power came from reaching every part of the factory and connecting separate machines into a system that could run continuously.
In the same way, asking AI to draft an email or summarize a report does not make a company AI-native. The deeper change begins when separate tasks are connected and AI can gather information, analyze a situation, take action, and adjust according to feedback.
The rise of agent systems such as OpenClaw has made this transition easier to imagine. Instead of only answering questions, these systems can use tools and turn an instruction into a series of actions under human authorization. Their development forces people who have built their careers around a narrow set of tasks to confront an uncomfortable question: if more execution moves to AI, where will my value come from?
The answer will not simply be “work harder.” A company’s advantage may increasingly depend on who can define the right problem, break complicated work into reliable processes, and understand where human intervention is still necessary.
AI Is Replacing Tasks and Expanding Access to Ability
If we speak only about replacement, AI begins to look like a force that does nothing but take work away. Yet it is also making knowledge, skills, and professional support more accessible.
In the past, gaining a new ability often required following a narrow path. A person might need to enter a particular school, find the right teacher, join an organization, pay for an expensive service, or spend years learning through trial and error.
Those paths still matter, especially when deep expertise is required. But they are no longer the only ways to begin.
Someone without a programming background can use AI to build a simple tool. A person with no formal design training can explore visual ideas. A beginner entering an unfamiliar subject can have an endlessly patient learning partner that adjusts its explanations and responds to questions whenever they arise.
AI is making some established skills less scarce while opening access to abilities that were previously difficult or expensive to acquire. It reduces not only the cost of completing a task, but also the cost of entering a new field and experimenting before deciding whether to go deeper.
Wang describes this as the democratization of services.
The Industrial Revolution made physical goods more widely available. Products that were once reserved for the wealthy entered ordinary homes through mass production. AI may do something similar for professional services, making basic forms of personalized teaching, health guidance, financial support, and expert assistance available to more people.
This idea reminded me of the AI weight-loss advisor I have been using.
In the past, receiving ongoing, personalized feedback about diet, exercise, and progress might have required hiring a nutritionist or personal trainer. AI cannot fully replace a doctor, nutritionist, or experienced coach, but it can help organize daily information, explain changes, respond to questions, and offer suggestions that reflect a person’s circumstances.
For me, its value is not simply that it generates a standard weight-loss plan. It gives me someone—or perhaps something—to consult when a specific question comes up. A process that might otherwise depend entirely on willpower now includes regular feedback and opportunities to adjust.
That is the less dramatic but more practical meaning of democratizing a service. It does not require pretending that AI already possesses the full ability of an experienced professional. It means that more people can receive a basic, continuous, and responsive form of support that was previously unavailable to them.
AI may therefore do for cognitive abilities and professional services what industrial production did for physical goods. It can give individuals access to support that once required an entire team or institution.
But opening the door does not mean everyone will walk through it, and those who do will not necessarily arrive at the same destination.
Wider Access Will Not End Inequality
The Industrial Revolution made goods more affordable, but it did not eliminate social divisions. It reorganized them. Some people owned factories and machines, some designed production systems, and others performed repetitive work on the assembly line.
AI may follow the same pattern. Easier access to knowledge and skills will not automatically produce equal outcomes. Instead, it may change the basis on which people become unequal.
Faced with the same technology, one person may use AI to complete assigned work more quickly. Another may use it to learn a new skill. Someone else may redesign an entire workflow or coordinate several agents to provide a complete service. Meanwhile, the companies that control models, platforms, data, and distribution will influence how everyone else can use these tools.
The emerging divide may be less about how much information a person possesses and more about whether they can formulate their own questions, judge the answers they receive, combine different tools, and turn an idea into sustained action.
As answers become easier to obtain, knowing what deserves to be asked becomes more valuable. As execution becomes cheaper, direction, judgment, taste, and responsibility become more important.
Still, it would be unfair to conclude that anyone who fails to benefit from AI simply did not try hard enough. People do not have equal amounts of time, education, money, freedom, or room to fail. Some can use AI to explore new possibilities, while others will be required to use it merely to produce more work. Some organizations will allow employees to redesign broken processes, while others will use AI to increase expectations without giving people greater autonomy.
AI lowers certain barriers, but it can also amplify advantages that already exist. What it democratizes is access to possibility, not a guaranteed result.
That may be the central tension of the AI era: more people will have the opportunity to become capable in ways that were previously unavailable to them, but the technology will not make everyone equally powerful.
Knowledge Can Spread Faster Than Experience
There is another reason wider access to AI will not erase the value of professional expertise.
Real expertise is rarely just a collection of facts or a set of steps that can be written in a textbook. It is judgment formed through practice, mistakes, feedback, and correction.
An experienced doctor may notice something unusual in a patient’s response. An engineer may recognize the early signs of failure from a subtle change in a machine’s sound. A teacher may sense that a student needs encouragement or a different explanation rather than more information.
We often describe such moments by saying that something “doesn’t feel right.” This is not necessarily a mysterious gift. It can be the compressed result of having encountered many situations that were similar but never exactly the same, having made mistakes, and having seen the consequences of different decisions.
AI can read medical textbooks, engineering manuals, and educational research, but it does not automatically possess all the feedback that professionals accumulate in the real world.
Whether AI can take over a professional task depends not only on the task’s intellectual difficulty, but also on whether the relevant experience can be recorded, whether actions produce clear feedback, whether success and failure can be measured, and who is responsible when something goes wrong.
Tasks that can be digitized, repeated, and evaluated clearly will be easier for AI to master. Work that depends heavily on physical presence, cultural context, human relationships, or personal responsibility will remain with people for longer.
Yet even this boundary may move. A single doctor or engineer can encounter only a limited number of cases in a lifetime, while an AI system may eventually learn from the accumulated experience of thousands of professionals.
Experience will not disappear, but its location may change. Instead of existing only in an individual’s memory and intuition, it may increasingly be distributed across systems built by people and AI together.
The professional of the future may create value not only by performing every action personally, but also by transferring experience into a system, recognizing what the system still fails to understand, intervening when something unusual happens, and accepting responsibility for consequences a machine cannot own.
The Bubble Comes Back to Us
AI can expand what people are capable of doing, but it cannot determine how those abilities will be used.
The same tool can help one person repeat the past and another escape it. It can be used to produce more disposable content or to create something that was previously impossible. AI is not distributing one predetermined future. It is distributing more ways for people to participate in making the future.
This brings me back to Wang’s prediction.
Technology does not become euphoric, and it does not become disappointed. People do.
We see AI complete several impressive tasks and assume that adoption will happen everywhere at once. We see a small number of people become extraordinarily capable with AI and imagine that everyone will receive the same result automatically. We see the possibility of profound long-term change and begin counting its economic value today.
But technical capability does not mean organizational adoption, adoption does not mean social transformation, and access to ability does not mean that everyone will use it.
When reality fails to arrive at the speed of our imagination, excitement turns into disappointment. What expands and bursts between those two emotional extremes is not the technology itself, but our expectations of it.
That is why the question of whether the AI bubble will burst in 2029 matters less to me than what happens before and after it.
As more routine execution moves to machines and more capabilities become widely accessible, we will have to reconsider where our own value comes from. Perhaps the scarce ability of the future will not be completing a task that someone else has already defined. It will be knowing which problems are worth solving, developing the experience to judge uncertain situations, and turning newly accessible capabilities into deliberate action.
AI can give more people the possibility of becoming powerful. How we become powerful, why we want that power, and what we ultimately choose to do with it remain human questions.