Chapter 015　Will AI Take Our Jobs?

of AI Will AI take our jobs? This is the most emotional question in the series. It is not a question about share prices, and it is not a question about strategy. It is a question about whether a person can go to work tomorrow. So the answer must not be hurried. “New jobs will appear, so we will be fine” is not an answer. It is the avoidance of one. In this chapter we establish what actually happened in earlier waves of technical change, identify where AI differs, and land on the responsibility the executive carries.

1 The question — why it arises now

This question has two hundred years of prehistory. In early nineteenth-century England, craftsmen broke the looms. History remembers them as the Luddites. They did not hate machines. They were protesting the fact that their skill was losing its price. Economically the protest was correct. The handloom weaver did vanish as an occupation. Employment in textiles then rose. Mechanization made cloth cheap, demand expanded, and factory work appeared in volume. In the aggregate, technology did not reduce employment. The textbooks have repeated this ever since. Two facts are both true at once. Employment across the industry rose, and particular skills became worthless. The people who took the new jobs were usually not the people who had lost the old ones. They were the next generation, and people in other regions. The twentieth century repeated the structure. The telephone operator disappeared as automatic switching spread. Assembly work moved to industrial robots. Clerical work was absorbed by the spreadsheet. Each time, new work appeared. Each time, some people could not move to it. Is what is happening now the next entry in that sequence? Or is it something else? The question takes this form now because the object has changed. The industrial revolution substituted for physical capability. Digitization substituted for information processing. What AI is substituting for is judgment and generation. Writing prose. Drawing a design. Assembling code. Forming a hypothesis. All of these sat on the human side for a long time. There is one more reason. The people who ask this question most urgently are not the people who decide the answer. Executives decide the answer. That asymmetry is what makes the question heavy. We therefore do not treat this as a question of forecasting. We treat it as a question of responsibility.

2 Conventional answers and their limits

Three answers circulate today. Each is partly correct. None is sufficient. The first answer: “Technology has never reduced employment, so we will be fine again” This is the most widely repeated answer. As a matter of record, no case since the industrial revolution shows technology reducing total employment over the long run. Productivity rises, prices fall, demand grows, and new work appears. The cycle has been observed many times. The answer has a limit. It confuses the aggregate with the individual. “Total employment did not fall” is a statement about a national accounting figure. Aggregates hide individual experience inside an average. When a skilled worker of forty loses a job and a twentyyear-old takes a new one, the statistics record employment as maintained. To the person who lost the job, that is no comfort at all. The second limit is the time axis. The adjustments history shows us often took a generation. The proposition that markets equilibrate in the long run is correct. During that long run, however, people have to live. What an executive faces is not the state after equilibrium. It is the few years before it. An executive who adopts this conventional answer overlooks the fact that somebody is paying the cost of the transition. That somebody is usually their own employee. The second answer: “AI will take our jobs, and mass unemployment is coming” The opposite answer circulates just as widely. Particular occupations are named, and their disappearance is predicted. The limit here is different. This answer counts occupations. An occupation is not a single solid object. Accounting exists as an occupation; a single task called accounting does not. Matching vouchers, posting entries, closing the month, handling tax, explaining results to the executive team, negotiating with the auditor. Dozens of distinct tasks are bundled together. AI can carry part of the bundle. When part is substituted, the occupation does not disappear. Its contents are exchanged. Historically this is the change that has happened most often. When automated teller machines arrived in banks, the branch staff did not vanish. Handing over cash shrank as a task, and advising and proposing grew. Predictions of the form “this occupation will disappear” are therefore usually wrong. They are wrong in a consistent way. The occupation survives, and the capability it demands changes. People who cannot move with the change lose their jobs individually. It appears as individual attrition rather than collective extinction. That is harder to address than collective extinction, because nobody comes to help. The third answer: “More new jobs will be created than AI takes away” The third answer sits on the optimistic side. Many of the occupations that exist today did not exist a century ago. So this time too, work without a name yet will appear. The claim itself is probably right. It cannot be used as a management answer. There are three reasons. First, the timing of loss and the timing of creation do not line up. Job loss comes first. New work comes later. The people in between absorb the interval on their own. Second, the person who loses and the person who gains are not necessarily the same person. New work usually demands new capability. Nothing guarantees that a clerical worker in their forties moves straight into a new occupation. Third, claims about the future require a reservation. No one can state accurately, now, what will be created. An expectation that something will appear cannot be placed where a measure should be. The three conventional answers share one gap. Each either files AI under the same heading as past technologies, or fears it as something wholly separate. What is needed is the specific list of where it is the same and where it is different. Three differences from earlier waves of technical change Three differences can be read from the published nature of the technology. None of them settles what will happen. As structural differences they are clear. First, the speed of diffusion. The steam engine and the industrial robot both required capital expenditure and physical installation. Adoption ran in years. AI is distributed mainly as software. It reaches a company with no equipment of its own the day after the contract is signed. The grace period for adaptation is structurally short. Second, the distribution of impact. Automation so far has thinned out routine work at intermediate skill levels. High-skill judgment work was held to be relatively safe. What AI is substituting for includes precisely that territory of judgment and generation. The impact reaches highly educated and highly paid groups, and that is new. Third, the granularity of substitution. Earlier machines replaced whole processes. AI replaces parts of processes, across every occupation. It moves broadly, shallowly, and simultaneously. So the question “which industries will be affected” is not framed correctly to begin with. These three points support neither optimism nor pessimism. They mean that the cost of transition can fall in a shorter period, across a wider population, all at once.

3 Redefinition — what is taken is not work; what is

returned is time Future Value Theory restates the question as follows.

Figure II-2　. Future Time — where the time AI returns actually goes

AI does not take work away. AI returns time. The question is who spends the returned time, and on what. This restatement is not wordplay. It moves the object of managerial decision to an entirely different place. Recasting work as a bundle of tasks The starting point is a change of unit. Stop counting work as occupations. Treat it as a bundle of tasks. Every job is made of twenty to fifty distinct activities. For each one, separate three cases. What AI can carry alone. What people and AI carry together. What people keep carrying. Run this decomposition and the picture changes. First, almost no job disappears whole. Many jobs will see 30 to 50 percent of the bundle replaced. Jobs where the whole bundle is replaced are rare. Second, the tasks that remain are strongly skewed in character. What remains is defining, choosing, owning the result, and handling exceptions. What goes is collecting, tidying, transcribing, and repeating a standard judgment. Third, the job does not reorganize itself. When 30 percent of a bundle empties out, whether the job takes a new shape is a matter of design. Without design, the empty 30 percent fills with other miscellaneous work. Or 30 percent of the headcount is cut. What AI decides, then, extends only to which tasks are replaced. What happens to a given person’s job is decided by the executive, not by AI. The decomposition has a second effect. When employees break their own jobs into tasks, the shape of their anxiety changes. The vague fear of “will my job disappear” becomes the concrete “these three activities will be replaced and these four will remain.” Concrete anxiety can be addressed. Vague anxiety cannot. Doing the decomposition and showing it is itself the work of management. The shift in the question — is it work that is taken, or time? Here we move the argument. The greatest value of AI is not the reduction of labor cost. It is the return of time. We call this Time Redefinition. Until now, time has been a constraint on the enterprise. Headcount multiplied by working hours set the ceiling on what a company could do. Growth therefore required hiring, and hiring required funds. AI loosens that constraint. Time spent preparing materials, time spent gathering information, time spent making routine judgments. All of it compresses. The compressed portion goes somewhere. There are only three destinations. The first destination is reduction. Convert the freed hours into headcount and subtract them from cost. Financial indicators improve. It is the fastest decision available and the least examined. The second destination is absorption. The freed hours quietly fill with other activity. Meetings increase. Reporting increases. Nothing changes except that the busyness is maintained. In most companies this is what actually happens. The third destination is the future. Time with customers. Time to learn. Time to think about a business that has no revenue yet. Only time allocated to this destination is called Future Time. Of the three, only the third changes the company’s future. And the destination is set by a managerial decision. Left alone, time always flows to the first or the second. People are capital — where Human Capital sits One more concept has to be redefined. Human Capital. Capital in Future Value Theory does not mean financial capital alone. Knowledge, people, data, trust, brand, networks, and AI. All of these are capital. People constitute one of them. This placement changes what a workforce reduction means. In accounting, personnel cost is an expense. Cutting expense raises profit. Seen through Future Value Theory, cutting people is cutting capital. Capital that has been cut does not appear in the financial statements. It appears years later. A new business is launched and no one can carry it. No one notices a shift in customers. No one notices an anomaly. Only then does it become clear what was lost. First Principle 4 sits at the root of this structure. AI Optimizes. Humans Define. AI optimizes; humans define value, purpose, and direction. However far AI advances, it cannot decide what should be aimed at. Choosing a purpose is choosing a meaning, and choosing a meaning is taking on responsibility. Only a human being can take on responsibility. The human role therefore does not shrink as AI performance rises. Whether the organization still holds the people who can carry that role is a separate matter.

4 Structure — what the returned time is multiplied by

Two equations carry the redefinition into practice.

4.1 The Future Time Equation

The fifth equation of Future Value Theory reads as follows. Future Value = Future Time × Future Capability We call it the Future Time Equation. The point to notice is that it is multiplication. It is not addition. Under addition, weakness in one term can be offset by the other. Under multiplication, the moment one term reaches zero, the whole product is zero. That property bears directly on the employment argument. When Future Time is zero. A company may hold capable people. If all of their hours are consumed by existing work, no Future Value appears. If AI freed up time and that time went to reduction or absorption, Future Time is still zero. When Future Capability is zero. Time can be returned. If the organization lacks the capability to use it, again no Future Value appears. Give free hours to an organization with no habit of learning, and all that grows is free hours. An AI deployment therefore means something only when two conditions hold together. Time is allocated to the future. And the people who can use that time are still in the company. A workforce reduction cuts both sides of this equation at once. The total stock of time falls, and the total stock of capability falls. The financials improve while Future Value is damaged twice over. When this structure is invisible, the reduction looks like a rational judgment.

4.2 Where Human sits in the Future Capital Equation

A second equation reinforces the point. Future Capital = Financial × Human × Learning × Trust × AI × Knowledge × Ecosystem × Purpose Eight terms, multiplied. Human is one of them. What the equation states is simple. If Human reaches zero, Future Capital is zero however large the other seven terms are. No amount of AI deployment compensates. No accumulation of Financial compensates. An abundance of financial capital cannot compensate for absent purpose, and advanced AI cannot compensate for absent trust. The same equation carries a warning in the opposite direction. Protecting Human alone gives the same result if Learning is zero. Holding on to people without supporting their relearning is not protecting capital. Protecting employment does not mean maintaining headcount. It means maintaining a state in which capability keeps being renewed. In practice the distinction is decisive.

4.3 Time in the Value Equation

Third, confirm the foundational equation. Value = Purpose × Trust × Capability × Time Time is an independent term here as well. With purpose, with trust, and with capability, value still does not appear without time. Because the relationship is multiplicative, value without purpose has no direction, without trust cannot spread through society, without capability cannot be realized, and without time cannot endure. In many companies this term is chronically short. Executives, managers, and the front line all spend their hours entirely on existing work. That is exactly why the time AI returns has a value of its own. It is something that could not previously be bought. Note the Trust term. A workforce reduction acts on it directly. Trust compounds faster than capital and becomes the last durable advantage, as First Principle 8 states. Speed works in the other direction too.

5 What it looks like in practice — where the pain

appears during the transition Return the structural argument to the floor of the company. Claiming that the transition produces no pain is not honest. Pain occurs. Its distribution is not even. The skew has at least three axes. Skew by generation The skew that deserves the most attention appears among younger employees. At first this looks backwards. Younger people are more comfortable with AI, and their payback period on relearning is longer. Adaptation should be easier. There is a structural problem. The staircase by which younger people accumulate experience is the part most easily replaced by AI. Every occupation has entry-level work. Taking minutes. Drafting materials. Collecting and cleaning data. Drawing a simple design. Writing junior-level code. The value of each is low in itself. The value lies in what it teaches: the shape of the whole job. These entry tasks overlap precisely with what AI does best. On efficiency alone, there is no reason not to replace them. But once they are replaced, where do the people capable of judgment ten years from now come from? This is not a question about an individual’s adaptability. It is a question of organizational design. Whether the developmental staircase survives is decided by the executive. Mid-career and older employees face a different form of pain. Part of the expertise they accumulated over decades loses its price. The deeper the expertise, the harder the conversion. What is needed here is not encouragement. It is a concrete redeployment design, and a guarantee of livelihood over the period it takes. Skew by region The second axis is geography. The same occupation can be affected everywhere, and the size of the pain still differs by place. In a city there is alternative work nearby. In a region dependent on one industry there is not. People are asked to change occupation and change where they live at the same time. For the enterprise this is not somebody else’s problem. When the major employer in a region cuts headcount, the regional economy contracts. In a contracted region, that company’s customers also decline. Employment is a cost and a market at the same time. Skew by occupation The third axis is occupation. Impact appears earliest in routine clerical processing, first-line service desks, simple translation, and entry-level production work. What they have in common is clear inputs, clear outputs, and few exceptions. The group most often overlooked is middle management. A large share of a manager’s work is relaying information. Up, down, and sideways. AI compresses that relay function heavily. This does not mean middle managers become unnecessary. As relaying shrinks, what remains is judgment and responsibility. The content of the role is exchanged. In companies that do not redesign the job to match the exchange, managers become the layer nobody can describe. And then they become the target of the next reduction. What the executive should consider before “AI-driven restructuring” This is the core of the chapter. AI is deployed and hours are freed. Cut headcount and profit rises. The arithmetic is correct. It is dangerous because it is correct. Before deciding on a reduction, at least four questions have to be passed through. Question one. What exactly was substituted? Was it a task, a judgment, or a responsibility? If only a task was substituted, redesigning that person’s job is enough. There is no need to cut people. Counting heads without making this distinction is not analysis. It is rounding. Question two. Where did the returned time go? Before choosing reduction as the destination, was allocation to the future examined as a destination? If it was not, this is not a choice. It is drift along the default path. Question three. What does reacquisition cost? If the people cut today are needed again in three years, can the same people be hired? What is the price in the recruitment market? How long does it take to reach proficiency? Put this calculation in, and the economics of reduction often reverse. Question four. How much will be paid in trust? A reduction acts on the employees who remain. In an organization that has understood “the gains from efficiency are used for cuts,” the front line will not cooperate with the next AI deployment. Improvement proposals stop. A decline in trust has no line item in the financial statements. So it is invisible. Being invisible does not stop it from happening. Even after all four, a reduction is sometimes necessary. Business environments do change structurally. What is at issue then is not whether to reduce, but how to reduce. The notice period. The redeployment options. Placement support. Opportunities to relearn. Designing these is also management. Two pictures of what it looks like in practice Place two companies side by side. In both, AI has cut hours by 20 percent. One cut 20 percent of its headcount. The remaining employees run the same work with AI. Financial indicators improved. A year later, not one new business has appeared. In the language of the Enterprise Redefinition Maturity Model (ERMM), this is the classic Level 2 state, the Improvement Enterprise. The source paper describes that state directly: organizations at this level become increasingly efficient while remaining fundamentally unchanged. Three qualifications travel with the ERMM and must be kept in view. Progression is not linear: organizations frequently display characteristics from multiple levels simultaneously, so a firm may hold Level 4 AI capability while remaining Level 2 in leadership. Maturity is assessed across all five dimensions in balance, because exceptional technological capability with weak leadership redesign cannot produce higher maturity. And Level 5 is not a target to be reached as rapidly as possible; different industries may require different levels of organizational adaptability. The other company kept its headcount. It allocated the freed 20 percent of hours to customer visits, to testing new territory, and to learning. Its short-term margin is below the first company’s. A year later, several businesses under test are running. What it looks like in three years is not yet known. We place no assertion there. The structural point can be stated. When the first company needs a new business in three years, neither the people nor the time to carry it will be inside the company. The difference between these two companies is not a difference in AI performance. The model deployed and the money spent can be almost identical. The difference was created in one meeting, where the destination of the freed time was decided. The technology arrives in the same form at the same price. Only what happens after it arrives differs from company to company. This is where the employment question connects to the management question. A company that can allocate returned time to the future has little need to cut people. New work is created inside it. A company that cannot design the destination of time has no option other than reduction. The power to protect employment comes from the quality of the design, not from the quantity of goodwill.

6 Questions for the executive

The argument, in one line. AI does not take work away. AI returns time. Whether that time goes to reduction or to the future is decided by the executive, and that decision decides the employment outcome. The phrase “AI took the jobs” deserves care. Putting technology in the subject position erases the location of responsibility. The decision to cut people is not made by AI. It is made by human beings. At the same time, we must not pretend there is no pain. During the transition, expertise is lost. Experience loses its price. People fail to keep up and fall out. That is hard to avoid. Precisely because it is hard to avoid, the question of who bears the cost of the adjustment matters. Three questions to close. They are not abstract. Each can be answered at your next executive meeting. Question 1 — Where did the time AI returned this year actually go? Most companies cannot answer. The effect of deployment is booked as hours removed, and the record ends there. Time has no ledger of destinations. Without a ledger, time is always absorbed by existing work. What share of the returned time became Future Time? That is a number you can measure. Question 2 — Is the staircase by which younger people gain experience still standing? Have you replaced entry-level tasks on efficiency alone? Where are the people who will carry judgment ten years from now being formed today? A company that cannot answer is taking profit now and selling its future capability. Question 3 — Before considering a reduction, did you design the redeployment? Designing redeployment does not mean handing surplus people some other activity. It means writing out, by name, the tasks that person keeps, the tasks that person takes on, and the learning required to get there. If it cannot be written out, the examination is not finished. None of the three questions asks what AI will take. All three ask what we will choose. AI returns time. People choose how it is used. Enterprises create the future. Society receives the result. The employment outcome is not decided by technology. It is decided by the sum of countless managerial judgments. Tomorrow, you will make one of them. Management in the Age of AI begins with taking on that responsibility.

In brief

AI does not take work away; it returns time. The question is what that time is used for.
Work is a bundle of tasks. Jobs that disappear whole are rare, but the tasks that remain are strongly skewed.
Returned time has three destinations: reduction, absorption, and the future. Without design it flows to the first two.
Cutting people is not cutting cost. It is cutting capital, and the effect appears quietly, years later.

Key concepts

Future Time / Future Capital / Future Value

The chain of ideas

AI → Future Time → Future Capital → Future Value

Related first principles

Principle 4 — AI Optimizes. Humans Define. Principle 3 — Capital Exists to Create Possibility. Principle 5 — Learning Is the Ultimate Competitive Advantage. Principle 8 — Trust Compounds Faster Than Capital.

Related chapters

Vol. I, Ch. 009 “How Should Work Be Divided Between AI and People?” — the principle for dividing task by task
Vol. II, Ch. 016 “What Kind of People Does the Age of AI Need?” — designing the returned time into renewed capability
Vol. II, Ch. 011 “What Is an Organization in the Age of AI?” — connecting the emptied processes to a redrawing of boundaries
Vol. V, Ch. 048 “What Does It Mean to Redefine Talent?” — recasting people as capital rather than cost

Papers and companion volumes

Kadowaki, N. (2026a). Future Value Theory: A Management Framework for Enterprise, Capital, and Society in the Age of AI. VURA Working Paper Series. SSRN: https://ssrn.com/abstract=7120980 / Zenodo: https://doi.org/10.5281/zenodo. 21255662
100 Questions on Management in the Age of AI, #001 “What Separates the People AI Displaces from the People It Does Not” / #046 “What the Executive Should Consider Before AIDriven Restructuring” / #047 “What You Do with the Time AI Frees Up Decides the Company”

Chapter 015 Will AI Take Our Jobs?