Chapter 016　What Kind of People Does the Age of AI Need?

of AI What kind of people does the Age of AI need? Most companies answer: people who can use AI well. The answer is not wrong. Its shelf life is far too short. A specification for people is meant to be a standard for selecting and developing them over a decade. A standard rewritten every few months does not function as a standard. In this chapter we stop defining people from an inventory of capabilities. We define them instead by the speed at which capability is renewed. Then we show how hiring, appraisal, and development have to be rebuilt to match.

1 The question — why it arises now

The specification for people has been rewritten in every era. In the industrial era, what was demanded was discipline and craft. Executing a fixed procedure accurately and repeatedly. That was a person’s value. In the information era, the demand moved to information processing and analysis. Reading documents, interpreting numbers, and turning them into reports. In the digital era, data and digital literacy were added. One idea runs through all three eras. List the capabilities required, and gather the people who hold that inventory. Job descriptions, certification schemes, and grading tables all stand on this idea. The idea worked for a long time. The reason is simple. Capability went obsolete more slowly than a career lasted. Expertise acquired in one’s twenties still held in one’s forties. So measuring the inventory made sense. That assumption is now breaking in three directions. First, the obsolescence cycle has become shorter than a career. The substance of the skills an occupation requires can turn over substantially in five years. The half-life of what we have been calling expertise has begun to fall below the length of a working life. Second, AI has taken on routine knowledge work. Looking things up, summarizing, drafting, writing the first version of code, cleaning up numbers. These have long been the work of junior people, and at the same time the foundation of expertise. That foundation is being externalized. Third, the developmental staircase is close to broken. When entry-level work moves to AI, the route from junior to mid-level disappears. Companies feel no pain for the moment. The pain comes after today’s mid-level people leave. Chapter 015 treated the question of whether AI takes work as a question about the structure of employment as a whole. This chapter treats the inside of it. What qualities are required of an individual? And how should the people function be designed to produce them?

2 Conventional answers and their limits

Three answers circulate. Each is partly correct. None functions as a specification for people. The first answer: “We need people who can use AI well” This is the most widely repeated answer. Many companies have written it into hiring criteria and appraisal items. Using generative AI daily. Building AI into the workflow. Setting out that requirement is not a mistake in itself. It does not last, for three reasons. First, proficiency with a tool has a short half-life. A way of writing instructions that worked in one period stops being needed when the model is updated. What we took for a skill is absorbed into the product as a feature in the next release. Second, operation gets democratized. Use of AI will follow the same path as use of any business system. In a few years, being able to do it will not be a differentiator, and being unable to do it will be a disqualification. A requirement that no longer separates people is useless for selection. Third, “can use it” cannot be measured. Frequency of use can be measured. Frequency and results are not proportional. A person who delegates even the judgments that must not be delegated looks excellent on frequency alone. The requirement should not be that a person can use AI. It should be that a person can doubt AI’s output. We make this concrete later in the chapter. The second answer: “Define the specification on a skills basis” The second answer is more sophisticated. Drop proxies such as education and employment history, and define the required skills by decomposing them. Build a skills map, make holdings visible, and fill the gaps with reskilling. As a piece of people practice this is clearly a step forward. In the Age of AI the method goes obsolete structurally. The reason lies in the unit itself. A skill is a name given to an activity that recurs. Market research, financial modeling, document preparation, coding. Each got a name because the same shape of activity appeared over and over. Because it had a name, it could be taught, measured, and priced. What AI takes on is exactly that recurring activity. When the activity is externalized, the thing the name pointed to disappears. The skill names stay in the personnel system and the contents drain out. The renewal speeds do not match either. Taking stock against a skills map takes a large company half a year. Define, survey, aggregate, approve. During those six months the meaning of the skills being surveyed changes. Being obsolete at the moment of completion becomes a permanent condition. The problem is not the quality of the execution. The problem is that measuring an inventory is structurally late. The third answer: “We need people with distinctly human capabilities” The third answer names creativity, empathy, ethics, and dialogue. It argues for placing people where AI cannot substitute. The direction is right. As a specification it has two weaknesses. First, the boundary moves every year. Part of what was once treated as the territory of creativity is already performed by AI at a high level. Write a requirement on top of a boundary, and the requirement moves when the boundary does. Second, this is definition by subtraction. Subtract what AI can do and assign the remainder to people. Under this idea, people are defined smaller every year. As a design philosophy for development, the direction is reversed. These words also cannot be measured in an appraisal setting. Told to rate creativity on a five-point scale, a manager ends up translating an impression into a number. The requirement becomes another name for subjectivity. The three conventional answers share one defect. They treat capability as a static inventory. They ask what a person holds, and never ask how long they can hold it. In an era of slow obsolescence, that was fine. It is not fine now.

3 Redefinition — what is needed is not an inventory of

capability but the speed of its renewal Future Value Theory defines the people the Age of AI needs as follows. The people who are needed are the people who keep renewing their capability. Not what they hold. What they discard, what they exchange it for, and how fast they can repeat the cycle. The value of a person lives in renewal speed, not in inventory. Why speed beats inventory Set up a simple thought experiment. Two people. One holds high expertise today. The other is somewhat behind on expertise, and relearns at twice the speed. After a year, the gap has not closed. After three years, the second catches up. After five, the second is ahead. What matters is that the reversal continues afterward. A difference in speed compounds. A difference in inventory erodes with time. In an era of slow obsolescence, retirement arrived before the reversal did. So measuring inventory was rational. In an era of fast obsolescence, the reversal happens inside a career. So measuring speed becomes rational. This is not an appeal to attitude. It is a change in what is assessed — an institutional judgment about what we agree to call the value of a person. Renewal speed decomposes into three qualities Renewal speed is hard to observe directly. Decomposed into three qualities, it leaves traces. First, the power to form a question. Learning is not the collection of answers. It begins by deciding what should be learned. AI returns answers to questions. No answer exceeds the quality of the question. This mirrors Question Design at the level of the enterprise (Kadowaki, 2026a). At the level of an individual, the quality of the question sets the quality of the learning. Second, the power to discard an assumption. Renewal is not addition. It is the act of letting go of an old assumption and putting a new one in its place. The main reason people cannot relearn is not a lack of time. It is that past success refuses to be discarded. The longer a person has delivered results, the stronger the resistance. Third, the power to connect to a purpose. Renewal hurts. There is a period of abandoning a familiar method and being incompetent again. What carries a person through that is not a sense of crisis. A sense of crisis works only in the short run and wears people down over the long run. What lasts is a purpose — an answer to what the renewal is for. Renewal without direction ends as pure attrition. The three are not independent. Forming a question requires doubting an assumption, and doubting assumptions continuously requires a purpose. Why the value of people rises as AI advances A common misreading should be corrected here: that the more knowledge work AI takes on, the less people are worth. Future Value Theory states the opposite. The more routine knowledge work AI takes on, the more people are worth. There are three reasons. First, the character of the remaining work changes. When activity is externalized, what stays in a person’s hands is defining, choosing, and owning the result. These are not merely non-substitutable. They are highly sensitive to enterprise outcomes. At the same job grade, the weight of the judgment carried rises. Second, evaluating AI’s output requires a human standard. Only someone who understands the domain can see an error inside an output. As AI performance rises, errors become subtler, and the standard required to catch them does not fall. It rises. Third, the results different people draw from the same AI vary widely. The model is common. The price is nearly common. The results still do not converge. What creates the difference is judgment: what to ask, where to doubt, and where to take the result on yourself. First Principle 4 states it. AI Optimizes. Humans Define. AI optimizes; humans define value, purpose, and direction. The quality of the defining side sets the ceiling on what optimization can deliver. Renewal speed is not an individual quality alone One more point is decisive. Renewal speed is not determined inside the individual. The same person keeps learning at one company and stops at another. This is observed daily. In an organization where doubting assumptions brings disadvantage, people do not doubt assumptions. In an organization where failure is penalized, people do not try new methods. In an organization with a six-month appraisal window, people attempt only what pays off in six months. Renewal speed is therefore the product of individual quality and organizational design. So this chapter cannot end with a discussion of qualities. It has to reach the design of the people function.

4 Structure — three forms of capital and two equations

To carry the redefinition into practice, connect it to the canonical structure.

4.1 Human and Learning in the Future Capital Equation

Future Value Theory reads the capital of an enterprise through this equation. Future Capital = Financial × Human × Learning × Trust × AI × Knowledge × Ecosystem × Purpose Eight terms, all joined by multiplication. Not addition. If one term is zero, the whole is zero. However thick the financial capital, if Human is zero then Future Capital is zero. An abundance of financial capital cannot compensate for absent purpose, and advanced AI cannot compensate for absent trust. What deserves attention here is that Human and Learning stand as separate terms. Having people, and having those people keep learning, are counted as two different forms of capital. Treating people as an inventory is looking only at Human. How many capable people do we hold? The equation says that is not enough. If Learning is zero, the product is zero however large Human grows. AI is a term in the same equation. Deploying AI is the act of enlarging one term. Enlarge that term alone and leave the others untouched, and the product does not grow as expected. When we are disappointed by the effect of a deployment, the cause usually sits in another term.

4.2 Where Learning sits in FVCC

The capability of an enterprise to create Future Value is expressed as follows. FVCC = Purpose × Learning × Redefinition × AI Integration × Ecosystem × Capital Allocation × Trust Learning is an independent term here too. The order carries meaning. Learning follows Purpose, and Redefinition follows Learning. Because there is a purpose, we learn; because we have learned, we can redefine ourselves. The order matches the Future Value Chain. Purpose → Learning → Redefinition → Creation → Enterprise Value The relationship is multiplicative: weakness in any single capability weakens the whole, so purpose, AI, and capital are each individually insufficient, and Future Value emerges only when all seven reinforce one another. The important point is that Learning here means organizational learning. It is not the simple sum of individual learning. The term grows only where a mechanism exists for what individuals learn to remain in the organization. In an organization where learning leaves when the learner leaves, Learning does not accumulate. In an organization where learning lives only inside individual heads, the same failure repeats in department after department. Raising individual renewal speed and embedding it in the organization require two different designs.

4.3 Why the Purpose term bears on the specification for people

Purpose appears in both equations. As Vol. II, Ch. 013 set out, Purpose governs culture. In the context of people, Purpose is equally decisive. There are two reasons. One is the direction-setting described above. Renewal that has no answer to what it is for does not continue. The other is not the logic of the side doing the selecting but the logic of the side being selected. People with high renewal speed can work anywhere. So their criterion for choosing a company is not compensation alone. What can I learn here? What is this company trying to achieve? This is the point taken up in 100 Questions on Management in the Age of AI, #043. Companies with weak Purpose lose their fastest-renewing people first. That loss does not appear in the current period’s income statement.

4.4 What Principle 5 means

The fifth First Principle states this whole chapter in one line. Learning Is the Ultimate Competitive Advantage. Learning is the ultimate competitive advantage, because knowledge and technology depreciate. Why “ultimate”? Because every other advantage can be copied. Equipment can be bought. Patents expire. People can be hired away. AI is available to everyone in the same form. What cannot be copied is the capability to keep learning. It is not an asset. It is a speed. And an enterprise’s learning capability stands on the learning capability of its people. That is why a discussion of people is a discussion of management. The principle carries an inconvenient implication for the executive. Learning is hard to measure and slow to show effect. So cutting it causes no pain in the current period. Investment in people is cut first because of that asymmetry. What is lost at the moment of the cut is not an expense. It is a competitive advantage several years out.

5 What it looks like in practice — rebuilding hiring,

appraisal, and development Take the redefinition into the three functions of the people organization. This is the practical center of the chapter.

5.1 Hiring — look for traces of renewal, not for inventory

Renewal speed cannot be measured directly. It leaves traces. Traces are what hiring should look for. We recommend four questions. What assumption have you discarded in the past three years? Not what you newly learned. What you let go of. A candidate who cannot answer has only been adding. What have you learned on your own, and how? Ask about learning chosen by the candidate, not training arranged by an employer. The power to form a question shows in how the subject was chosen. What conclusion did you draw from a failure? Not whether there was a failure, but the generalization drawn from it. Renewal happens through the interpretation of failure. What did you delegate to AI, what did you keep, and why? This is the question that replaces “can you use AI.” Judgment shows not in the scope delegated but in the reason for not delegating. Asking about volume of use reveals nothing. Asking how the boundary was drawn reveals a great deal. Change how requirements are written as well. Stop listing skills and describe the questions the role must carry. What questions is this job responsible for answering, continuously? Skills change in service of the question; the question holds for years. The same principle reaches graduate hiring. Measuring readiness to perform on day one matters less than it used to, because skills at entry turn over within two years. What to look for is whether a way of learning has been established as a method inside the person.

5.2 Appraisal — put renewal beside results

If appraisal does not change, changing hiring means nothing. People do what they are appraised on. Most companies appraise on a single axis: results. That design works against renewal. Renewal involves a period in which results temporarily fall. Between discarding a familiar method and becoming proficient in a new one, productivity drops. On a single axis, that period is pure loss. Appraisal therefore needs a second axis. Results, and renewal. State concretely what the renewal axis looks at. Is there a documented case of rewriting an assumption? Was a new method tried and the outcome reported? Was the person’s learning turned into a form others can use? The third matters most. It is the act that converts individual learning into organizational Learning. The treatment of failure belongs to the design. A failure incurred to test an assumption and a failure caused by carelessness must not be treated alike. Penalize the first and renewal stops. As Vol. II, Ch. 013 argued, how failure is handled sets the speed of learning. The appraisal period needs review too. In an organization with a six-month window, only what pays off in six months gets tried. The renewal axis alone should be read on a longer cycle.

5.3 Development — how to rebuild a staircase that is breaking

This is the hardest problem, and the most overlooked. How have people developed until now? They processed entrylevel work in volume, acquired the fundamentals through it, and advanced to mid-level judgment. Someone who has produced dozens of documents becomes able to judge whether a document is good. Someone who has written hundreds of first drafts becomes able to fix another person’s first draft. In other words, entry-level work was a deliverable and a training device at the same time. It performed two functions at once. When AI takes over entry-level work, only the deliverable function is inherited. The training function is inherited by nobody. A structural break opens there. The effect is not immediate. Today’s mid-level people climbed the staircase before it broke. Operations continue for the time being. The problem surfaces five to ten years out. The mid-level people who can exercise judgment are not there in the numbers they should be. That hollow appears nowhere in the financial statements. The structure appears regardless of sector. Software development, legal, audit, consulting, engineering design in manufacturing. All of them built judgment through an apprenticeship of junior repetition. We propose four responses. First, separate training from operations and design it explicitly. Training used to be a by-product of the work. Now that the by-product is gone, it has to be made deliberately. This is a decision to buy deliberate inefficiency. Reserve budget for hours in which people perform, for the sake of their development, work that AI would finish faster. It is an expense. It is also the purchase of a midlevel cohort five years from now. Second, raise the level of the work assigned. Give junior people AI as their staff. Delegate the activity to AI and give the junior person the design and the verification. Start the practice of judgment in year three, where it used to start in year ten. If the training device is broken, redesign the entry point to a higher step. Third, put them on the reviewing side. Have them critique AI’s output, and have experienced people critique the critique. Judgment develops as much through detecting as through producing. AI supplies material without limit. This is one of the few opportunities created by the breaking of the staircase. Fourth, put the tacit knowledge of experienced people into words. Left as private intuition, experience leaves with the person. Put the grounds for judgment into language, record them, and turn them into teaching material. This thickens the Knowledge term in the Future Capital Equation. Experience does not become worthless. It simply does not become capital until it is put into words. This point is treated in 100 Questions on Management in the Age of AI, #045.

5.4 The three move as one

Hiring, appraisal, and development fail whenever they are touched separately. Hire for renewal speed while appraising on results alone, and the people hired burn out within a few years. Invest in development while appraising on a short cycle, and the people invested in lose out. Change appraisal alone while the requirements stay old, and there is nothing settled to appraise against. One standard can run through all three. Does renewing yourself get rewarded at this company? If the front line can answer yes, the details of the system will settle themselves later.

6 Questions for the executive

The argument, in one line. The people the Age of AI needs are not the people with the largest inventory of capability but the people who keep renewing capability toward a purpose, and the work of management is to design a system in which that renewal is rewarded. It is not a search for people who can use AI. It is not a search for people who are distinctly human. Those are parts of the design, no more. Three questions to close. Each can be answered at your next executive meeting. Question 1 — When was your specification for people written? Check the date on it. If it is more than three years old, the meaning of the skills written there has already changed. We are not arguing that requirements should be rewritten annually. Written as questions rather than skills, a requirement holds for years. The question is whether yours was written to hold. Question 2 — Is there any difference in how people who renewed and people who did not are treated? List the promotions of the past three years and check the stated reasons. If results are the only reason, this company is appraising inventory. If no one has been rewarded for rewriting an assumption, employees are behaving correctly. The system is choosing nonrenewal on their behalf. Question 3 — Where is your mid-level cohort of five years from now being formed today? If you cannot name the place, the staircase is already broken. Every piece of entry-level work AI absorbed removed an occasion for training. Check whether the design that fills that gap is in this year’s budget. None of the three questions asks whether your people are good. All three ask whether this company has a mechanism that keeps people renewed. People are capital. Capital left alone deteriorates. As equipment is depreciated, capability goes obsolete. The difference is that capability alone can appreciate, depending on the design. The third First Principle states the role of capital. Capital Exists to Create Possibility. Capital exists to create possibility, not merely to maximize return. Human capital is no exception. Investing in people does not mean retaining them. It means holding them in a state where they can create new possibility. AI substitutes for part of our capability. It also returns the time and the means for renewal. What that returned time gets multiplied by is the single point on which the future of an enterprise divides.

In brief

The people the Age of AI needs are not those with the largest inventory of capability but those who keep renewing it.
Renewal speed decomposes into three qualities: forming a question, discarding an assumption, and connecting to a purpose.
Renewal speed is the product of individual quality and organizational design, so hiring, appraisal, and development are rebuilt as one.
The more routine work AI carries, the more people are worth, because the quality of the defining side sets the ceiling on results.

Key concepts

Future Capital / Future Value Creation Capability / Question Design / Future Value

The chain of ideas

Purpose → Question Design → Learning → Future Capital → Future Value

Related first principles

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

Related chapters

Vol. II, Ch. 015 “Will AI Take Our Jobs?” — the relation between returned time and human capital
Vol. II, Ch. 018 “What Should a CEO Learn in the Age of AI?” — when the object of renewal is the executive
Vol. V, Ch. 048 “What Does It Mean to Redefine Talent?” — the redefinition of people turned into a working procedure
Vol. VII, Ch. 070 “Do People Become Enterprise Value?” — how far people can be spoken of as value

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, #043 “What Kind of Company Do Outstanding People Choose?” / #059 “What Should You Look For When Hiring in the Age of AI?” / #045 “Does the Experience of a Veteran Become Worthless in the Age of AI?”

Chapter 016 What Kind of People Does the Age of AI Need?