Chapter 011　What Is an Organization in the Age of AI?

of AI What is an organization in the Age of AI? The question is usually swapped for a smaller one: how should we redraw the org chart? Cut a layer. Merge two divisions. Decide where the AI unit reports. But an org chart is not an organization. It is one drawing of where people sit. What changes in a company that has brought AI inside is not where people sit. It is the mechanism by which value is produced. In this chapter we recast the organization from a collection of people into a value creation system.

1 The question — why it arises now

Twentieth-century organization theory was built on a single constraint. Information sits unevenly, and moving it costs money. We call the two halves of that constraint information asymmetry and transmission cost. Every organizational form was an answer to them. The bureaucracy Weber described is the first answer. Gather information upward, test it against rules, and push instructions down. Hierarchy was a device for dividing limited information-processing capacity among people. The line showing who reports to whom was also a plumbing diagram for information. The divisional structure is the next answer. Sloan established it in the automobile industry, and Chandler set it out as theory (Chandler, 1962). Once a company serves several markets, headquarters can no longer process the information from all of them. So decision rights were pushed down toward the places where the information already was. The divisional structure is a technique for splitting an organization along the limits of information processing. The matrix belongs to the same lineage. Product information and regional information had to be handled at once, so a second reporting line was drawn. Teal and self-managing models do not change the premise either. The front line holds the most information, so let the front line decide. The forms differ enormously. The premise does not. Information is unevenly distributed and expensive to move, so arrange people to minimize that expense. Organizational structure was, in essence, the design of information plumbing. The same logic governed the boundary of the firm. Coase (1937) explained that boundary through transaction costs, and Williamson extended the account. If buying in the market is expensive, hold the work inside. If it is cheap, push it out. The line of the employment contract was drawn as the result of a cost calculation. AI is now dissolving that premise. AI reads the whole company’s information at once. It holds context across departments. It translates specialist vocabulary, produces summaries, and delivers them to whoever needs them. The cost of moving information does not fall to zero. It changes by an order of magnitude. Transmission is not the only thing that changes. An AI agent is not a tool waiting for instructions. Given a purpose, it assembles its own sequence of steps across process boundaries. It is a recipient of information and a doer of work at the same time. Quietly, then, the number of workers inside the organization who hold no employment contract keeps rising. Two things have happened together. Transmission cost collapsed, and the population of doers diversified. That is why the organization is worth questioning now. Which raises the question. When the structure built to move information is no longer needed, what does the organization exist for? This is not a question about redrawing the chart. It is a question about the concept itself.

2 Conventional answers and their limits

Three answers circulate. Each is partly right. Each sees half of the organization. The first answer: “Organizations in the Age of AI go flat” This is the answer we hear most. Hierarchy existed to move information. If AI moves information, the middle layer is unnecessary. So the organization flattens. The logic holds. Meetings held to report, the aggregation of materials, the translation between departments — this work is losing its value quickly. But the answer has a hole in it. Hierarchy never carried transmission alone. It arbitrated priority. It allocated resources. It handled exceptions. It grew people. And it took on the consequences of what had been decided. None of that disappears when transmission gets cheap. An organization that makes flatness the objective sets out to cut the transmission function and cuts the arbitration function with it. Every judgment then flows back up to the executive team. The organization is flatter, and decisions are slower than before. We have watched this happen in many companies. Flatness can occur as the result of a design. It cannot be the objective. The second answer: “Break the company into small, fast teams” The second answer is about division. Create many small autonomous teams and give each one discretion. Software organizations have used this widely, and its effectiveness is demonstrated. Small units learn fast, because the distance from decision to result is short. We do not dispute that. But division is only half of a design. The smaller the units, the more gaps open between them. And in the Age of AI, much of the value arises not inside a unit but at the seam between units. New value stands up where knowledge from different domains connects. An organization that divides without designing connection has renamed departmental optimization. The same failure appears, in the same shape, in the context of AI deployment (→ Vol. I, Ch. 010). The third answer: “An organization is a group of people, and AI is its tool” The third is less a stated answer than an assumption most companies hold without noticing. HR counts people. The org chart arranges people. The budget stacks up personnel cost. AI is a tool, so it lands in a different column as an expense. Under that assumption something strange happens. The work an AI agent processes daily appears nowhere on the chart. Neither does the core process run by a contracted engineer. Neither does the design knowledge shared with a partner firm, nor the improvement signal arriving from customers every day. The structure that actually produces value and the structure management is looking at have come apart. Design an organization from a map that is off, and the design will be off. Headcount falls and the work does not. Departments merge and collaboration does not follow. The cause sits in the part that is not on the map. The three conventional answers share one limit. All three ask how to arrange people. What AI is putting to us is the question one level above that. What, connected to what, produces value?

3 Redefinition — the organization is a value creation

system Enterprise Redefinition recasts the enterprise across five dimensions: Purpose, Business, Organization, Capital, and Leadership. The third of them is Organization.

Figure II-1　. The organization as a value-creating system

The definition there is plain. An organization is not a collection of people. It is a value creation system in which people, AI, robots, external partners, startups, universities, and customers are joined into one. That single line changes three things at once. What the organization is made of, where its boundary lies, and how its capital is understood. We take them in order. Change 1: constituents — the roster does not match the employment contract First, what the organization is made of changes. Not only people. An AI agent carries part of a value-producing process. So does a robot on the production floor. So does the university laboratory running joint research, and so does the startup you have partnered with. And the customer who uses the product and returns information about improving it is, in substance, part of the system too. This is not a metaphor. If stopping a given process stops value creation, that process is a component of the system. Recount your own company on that criterion. The number of people holding an employment contract and the number of agents producing value will not match at all. What we have been calling the organization was in fact the set of employment contracts. Organizational design in the Age of AI begins by pulling those two apart. Change 2: the boundary — a gradient, not a line Second, the boundary changes. In many companies the following is already routine. The same stage of the same project is shared between an employee and a contracted specialist. A partner firm’s engineer joins, and so does an AI agent. Look at the output and you cannot tell which of them did what. To the customer it matters even less. The boundary has not vanished. It has changed from a line into a gradient. It is no longer one line dividing inside from outside. How far do we share information? How far do we admit others into the decision? How far do we share the result? The boundary is the degree of that permeability. The criterion for drawing it changes as a result. Inside and outside used to be decided by cost. If it can be made more cheaply outside, send it out. That is the transaction-cost logic. The more AI drives down the cost of coordination, the more that logic answers “send it out,” every time. A company that follows the answer to its end is hollow within a few years. So the criterion has to be replaced. What belongs inside is the process in which learning accumulates, and the process that carries responsibility. Those two do not come back once they leave. An outsourced process accumulates experience, but not your learning. A company that hands responsibility outward stops being the subject that judges. For every other process, the further the boundary opens outward, the stronger the value creation system becomes. Change 3: capital — a relationship with the outside is capital The third change is the one most often missed. The Future Capital Equation treats capital as the product of eight terms. Future Capital = Financial × Human × Learning × Trust × AI × Knowledge × Ecosystem × Purpose The seventh term is Ecosystem. In this chapter we call that term Ecosystem Capital: the accumulated stock of relationships with external agents through which a company can create value jointly. Ecosystem Capital is not a trading relationship. A transaction ends when the contract ends. Ecosystem Capital remains after the contract ends. With that laboratory, the conversation starts halfway in. With that startup, we can reach a prototype in three weeks. None of this appears in the financial statements, and all of it moves Future Value. And it is one term in a product. Because the relationship is multiplicative, a zero here makes the whole zero, however large the other seven terms are. An abundance of financial capital cannot compensate for absent purpose, and advanced AI cannot compensate for absent trust. A company that completes everything within its own walls is operating with this term left at zero. A stock is not value until something moves it. That requires capability. Look at the FVCC Formula. FVCC = Purpose × Learning × Redefinition × AI Integration × Ecosystem × Capital Allocation × Trust Ecosystem appears here as well. This one sits on the capability side, and we call it Ecosystem Capability. If Ecosystem Capital is the accumulation, Ecosystem Capability is its operation. Here too the relationship is multiplicative: weakness in any single capability weakens the whole, and Future Value emerges only when all seven reinforce one another. Ecosystem Capability has four parts. Finding the right counterpart. Translating the purposes of different organizations into shared language. Designing the split of the result in advance. And the discipline that keeps trust intact when the work does not go well. A company without the fourth can partner once. There is no second time. What middle management becomes When constituents, boundary, and capital all change, the role that moves furthest is middle management. Until now the middle manager was the junction of information. Translate intent downward, aggregate reality upward. The share of working hours spent on that is startlingly high in most companies. And AI is taking it over. So what does the role become? Answered from the standpoint of organizational design, it becomes the designer of a value creation unit. This is the person who designs their own territory as one small value creation system. What sits inside. Which external agents to connect to. Which steps to hand to AI. Where a person must judge. Corporate planning cannot do this design. Only someone who knows the reality of the front line can know where the boundary should run. The principle governing where responsibility sits inside that design is handled elsewhere (→ Vol. I, Ch. 009). What we are describing here is not responsibility. It is structure. Not who carries what, but what connects to what. This is not a shrinking of the role. It means the measure of a manager changes. Not the number of subordinates, but the value the territory produced and the degree to which that territory is connected outward.

4 Structure — designing what the org chart cannot show

Now carry the redefinition into practice. The limit of the org chart cannot be avoided here. An org chart can show three things. Who reports to whom, which departments exist, and how many people are in each. What it cannot show is longer. Where information flows. Who holds the decision right. How far the boundary is transparent. Where trust is thick and where it is thin. Where learning happens and where it accumulates. In the Age of AI the performance of an organization is decided almost entirely by the second list. So we need language for designing outside the chart.

4.1 Four objects of design

We organize the objects of organizational design into four. First, the distribution of purpose. This means deciding the question each unit is to answer. Not its assigned duties. Its question. “Protect revenue in the western region” is a duty. “What is still missing from this product for customers in the western region?” is a question. Distribute questions and units begin to think. Distribute duties and units wait for instructions. Second, the placement of decision rights. Make explicit who decides and who may overturn the decision. Now that information asymmetry has narrowed, the reason for holding authority at the top has narrowed with it. But unless someone is named as able to overturn, the organization will simply leave judgments unmade. Third, the design of connection. By what means is unit joined to unit, and inside joined to outside? By a meeting, by a shared record, or by an AI-generated summary? Division carried out without designing connection always becomes rupture. Fourth, the cycle of learning. Where is learning done, where is it deposited, and who can withdraw it? Learning stored in one person’s head leaves when that person leaves. Learning stored in the system stays. None of the four can be drawn on an org chart. All four are the substance of the organization.

4.2 Trust does the work that structure used to do

Apply the Value Equation to the organization. Value = Purpose × Trust × Capability × Time This is a product as well. If one term is zero, the whole is zero. 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. The position of Trust is what to notice. The further the boundary moves from a line toward a gradient, the more value is created with counterparts a contract cannot bind. The joint research partner. The allied startup. The customer who sends back improvement data. A contract does not control these relationships. Trust does. Order inside an organization used to come from structure. People moved because a chain of command existed. Where the boundary has blurred, trust takes over part of what structure carried. First Principle 8 states it. Trust Compounds Faster Than Capital. Trust compounds faster than capital and becomes the last durable advantage. And what grows fast is lost fast. Ecosystem Capital usually breaks through a single act of bad faith.

4.3 What an adaptable structure is

The Enterprise Redefinition Maturity Model (ERMM) puts one question to the Organization dimension. “Can structures adapt rapidly to technological change?” The question does not ask for the correct structure. It asks for a structure that can be rebuilt. Any form stops being optimal once the technology moves. So the test is not whether the form is good. The test is how long it takes to change the form. A company that needs a year to change its structure is permanently behind in an environment whose cycle is shorter than a year. A company that can rebuild within a quarter is not behind. ERMM Level 4 is the Continuous Redefinition Enterprise, and adaptive structures are listed among its characteristics in exactly this sense. What decides the speed of rebuilding is not the number of reporting lines. It is whether the four objects of design in the previous section have been written down. Where purpose, authority, connection, and the place learning is deposited all remain tacit, every attempt to rebuild starts the argument from nothing. Three cautions travel with the ERMM and must not be dropped. Progression is not linear: organizations frequently display characteristics from multiple levels simultaneously, and a firm may hold Level 4 AI capability while remaining Level 2 in leadership. The model evaluates organizational coherence rather than isolated excellence. Maturity is assessed across all five dimensions in balance, since exceptional technological capability with weak leadership redesign cannot produce higher maturity. And the objective is not reaching Level 5 as rapidly as possible; different industries may require different levels of organizational adaptability.

5 What it looks like in practice — redesigning as a value

creation system How is the redesign actually done? We set it out in order.

5.1 Seven steps

Step 1. Draw the flow of value. Put the org chart aside and draw the flow of value on a blank sheet. Where does value for the customer arise, and by what route does it reach them? Write no department names. Write only the steps. Step 2. Write the doer of each step, with titles hidden. Record who carries each step without recording employment status. A person, an AI, a machine, or an external agent. Most executives see their own company accurately for the first time here. Step 3. Identify the junctions that exist only for information. Some steps add no value and only pass information from left to right. Sift meetings, approvals, and reports on this basis. This is the part AI changes most. Step 4. Redraw the units. Cut units along blocks of the value flow rather than along departments. Size each unit so that it can answer one question. Step 5. Set the permeability of each boundary. For each unit, decide what stays inside and what opens outward. The criterion is the one above: does learning accumulate, and is responsibility carried? Step 6. Design connection and the cycle of learning. By what means are units joined, and inside joined to outside? Where does learning accumulate? Skip this step and the division from Step 4 turns into rupture. Step 7. Draw the org chart last. The chart is not the starting point of a design. It is the record of one. Whether this order is kept changes the result completely. Most companies begin at Step 7. Draw boxes, name them, fit people into them. That is a change of placement, not a design.

5.2 What it looks like across industries

Organizations already run as value creation systems. We describe them qualitatively, within what is publicly known. An advanced manufacturing plant is no longer a group of people. It is a mixed system of people, industrial robots, and AI carrying demand forecasting and anomaly detection. Engineers from the equipment maker are often resident on site, and improvement themes are shared with a local university. Knowledge that supposedly sits outside the organization is built into daily operations. In pharmaceuticals, drug discovery stopped being the work of a single company long ago. A corporate laboratory, university basic research, a biotech venture, and a firm running AI-based search for candidate compounds. That coalition functions as one research organization. Which member made the discovery matters less than the search speed of the coalition, which has become the unit of competition. In retail and consumer goods, the customer is part of the design. Reviews, usage data, and requests are design information for the next product. Some companies define the customer as a purchaser outside the organization. Others define the customer as a component of the value creation system. Product development speed differs structurally between the two. Financial services and professional services are moving the same way. Screening and research steps are increasingly shared among AI, outside specialists, and the firm’s own staff. What stays inside is the step that sets the criterion of judgment, and the step that carries responsibility for it. Not who performed the work, but who holds the criterion. That is what now draws the outline of a firm. Software offers a more extreme form. A core component of a company’s own product being maintained by developers who are not employees is an ordinary condition. Open source is a design that deliberately opens the organizational boundary outward.

5.3 What failure looks like

Designs that do not work share recognizable shapes. One is confining AI to a single department. Create a promotion unit and staff it with specialists. This is effective at the start. Let two years pass unchanged and AI becomes that department’s tool. It never becomes part of the value creation system. Another is redrawing only the org chart. Boxes get new names and reporting lines are redrawn. The flow of value has not changed at all. Within six months people are moving along the old flow again. The structure was supposedly changed, and in fact nothing was. A third is confusing the ecosystem with procurement. Treat partners as suppliers and squeeze terms for short-term gain. The financials improve. Ecosystem Capital quietly falls. And when the next hard attempt comes, nobody will take it on with you. A fourth is opening the boundary while keeping the information closed. The joint research contract is signed. The design information that matters is not shared. The counterpart can only work at the surface, and no result appears. What remains inside the company is the conclusion that partnering does not produce results. Permeability is not decided by the presence of a contract. It is decided by three things: information, involvement in decisions, and the distribution of the result. Every one of these failures has the same cause. The organization is still being treated as a question about how to arrange people.

6 Questions for the executive

The argument, in one line. An organization in the Age of AI is a value creation system: a mechanism joining people, AI, robots, external partners, startups, universities, and customers into one. It is not a collection of people. Rearranging people therefore does not change it. It changes when you redesign what is connected to what, where learning happens, and how far the boundary opens. First Principle 6 states it. Enterprise Exists to Redefine Itself. Continuous self-redefinition is the essence of the enterprise. The organization is the vessel through which that redefinition is carried out. Where the vessel stays rigid, no conception, however good, takes shape. Three questions to close. Each can be answered at your next executive meeting. Question 1 — What share of your value is produced by agents who are not on your org chart? AI agents, contracted specialists, partner firms, joint research counterparts, and customers. Recount with all of them included. What share of the value is produced by people holding an employment contract? If that share is below half, your org chart no longer explains your company. Question 2 — What are your middle managers designing right now? Examine how they spend their hours and you will know. If most of it goes into aggregating and relaying reports, the role is being replaced by AI. If it goes into designing the boundary and the connections of their territory, the role is getting heavier. The difference is not one of capability. It is a difference in what has been expected of them. Question 3 — Over the past year, did your boundary widen outward or contract inward? How many new external agents did you join with? How many joint attempts did you start? And how many relationships were closed off in the name of bringing work back in-house? Ecosystem Capital erodes without fail unless it is deliberately increased. None of the three questions asks about the shape of the organization. All three ask what the organization is connected to. The era of arranging people in order to move information is ending. The organization ahead is a design that connects agents in order to produce value. Those agents sit inside the company and outside it. They are not necessarily people. And what to connect to is not decided by AI. Connection carries responsibility, and only human beings can take responsibility on. Organizational design in the Age of AI is the work of an executive deciding again how far the enterprise counts as itself.

In brief

An organization in the Age of AI is not a collection of people. It is a value creation system joining people, AI, and external agents.
Rearranging people does not change an organization. What it is connected to, and how far its boundary opens, is what decides it.
Only two things belong inside: the process in which learning accumulates, and the process that carries responsibility. The rest can open outward.
Ecosystem is one term in a product. A company that completes everything within its own walls is operating with that term at zero.

Key concepts

Enterprise Redefinition / Future Capital / Future Value Creation Capability / Future Value

The chain of ideas

Enterprise Redefinition → Organization → Ecosystem → Future Capital → Future Value

Related first principles

Principle 6 — Enterprise Exists to Redefine Itself. Principle 8 — Trust Compounds Faster Than Capital. Principle 5 — Learning Is the Ultimate Competitive Advantage.

Related chapters

Vol. I, Ch. 009 “How Should Work Be Divided Between AI and People?” — the principle governing where responsibility sits is there, not here
Vol. II, Ch. 012 “What Is Leadership in the Age of AI?” — how distributed leadership moves an organization
Vol. V, Ch. 047 “What Does It Mean to Redefine the Organization?” — the redefinition of the organization set out as practical procedure
Vol. IV, Ch. 033 “Are Intangible Assets Future Value?” — the argument for measuring accumulated relationships as capital

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
Kadowaki, N. (2026b). Enterprise Redefinition: Toward an Enterprise Evolution Theory for the Age of AI. VURA Working Paper Series. (Published on Zenodo; under review at SSRN)
100 Questions on Management in the Age of AI, #041 “The Day Your Reports Become AI: What Happens to the Manager’s Job” / #057 “The Next Growth Is Outside the Company”

Chapter 011 What Is an Organization in the Age of AI?