In July 2004 the New York Times reported from the Milan men's wear shows that the posse of American editors and buyers in the front rows at Dolce & Gabbana, Gucci and Burberry, the men whose job is telling other men what to wear, had, to a man, worn their shirttails untucked. Seven years later a company called Untuckit was founded to sell shirts cut to be worn that way. In 2013 the chief executive of Apple appeared on stage at the company's developer conference with his shirt untucked. Google searches for untucked shirts peaked in 2017. By 2020 the Banana Republic factory outlets had a section labeled "untucked oxford shirts."
Fashion is the fruit fly of commerce: the generations are short, so you can watch the life cycle of a category in a decade. Every category goes through the same four phases, which I call sense, solve, scale and sustain. The alpha asset that wins in one phase is not the one that wins in the next. Sensing is the recognition of a disequilibrium: a need without a solution, or a new way to meet an old need. At this point the idea, which usually occurs to several people at once, is a hypothesis clouded by uncertainty. Solving is the scramble of rival attempts, most of which we never hear about, that lasts until one or a few find a fit between the solution and the need and, in hardware, a dominant design emerges. Scaling is when the survivors grow, the supply chain forms, the customer learns the category, and the flywheels of chapter 5 begin to turn. Sustaining is the long period, often decades, in which an oligopoly of survivors runs the flywheels and earns most of the category's profit. (A bell foundry in Italy has been sustaining since 1339.)
The most consequential transition is the one from solve to scale. During solve, the only alpha asset a new firm has is a solution better than the alternatives for some customers, plus the speed to get it to them. During scale and sustain, every surviving rival has a solution, and the assets are brand, scale, the customer network and capability. A firm that is good at one phase is usually bad at the next. The organization theorist James March named the general tension exploration and exploitation (looking for the oil versus extracting it) and observed that organizations optimized for one are poorly built for the other. The habits that run a sustaining business (key performance indicators, process control, structured hiring, global expansion) would have prevented the business from being founded. Chapter 7 is about what a founder should do during the solve phase. The prior question is where to be at all.
Fit
The principle is short: seek the setting in which the assets you have, or can build, are rare and necessary. It holds for a company choosing a market and for a person choosing a job, for the same reasons.
By the definition in chapter 2, an asset is an advantage only relative to the people competing for the same work in the same setting. One consequence is easy to state and hard to accept: you do not need to be the best in the world at anything. You need a combination of capabilities that is unusual among the people competing for this particular work, in a setting where that combination improves results. Economists call the general form comparative advantage: a person who is worse than a rival at two tasks can still be the right person for one of them, if the gap is smaller there. A mechanical engineer who writes clearly is an ordinary engineer in a room of engineers and an ordinary writer in a room of writers. In a room of people who must explain engineering decisions to executives, they have no competition.
Fit also runs the other way, because a setting supplies complements. The portability finding in chapter 4 was that much of a star's performance is the platform they stand on. The star securities analysts in Boris Groysberg's study lost rank when they moved to new firms, because the research platform, the company relationships and the colleagues stayed behind. So the question for a person is not only where their capabilities are scarce but where their complements are present, and who owns them. A firm supplies capital, customers, reputation and colleagues; the person supplies judgment and effort. How the returns are divided, as Russell Coff observed, is a matter of bargaining power. The engineer whose results depend on the firm's platform has less of it than they think; the engineer who holds the relationships has more.
For a Company
The five flywheels of chapter 5 are built in one market and can then be used to enter another. Which directions are open depends on where the firm's alpha assets reach.
Geographic expansion usually comes first. The jobs to be done are similar everywhere; everyone eats, moves, communicates and wants to be entertained. What differs is the friction: regulation, logistics, language, taste. My coauthor Lele Sang and I studied eight multinational companies entering China and found a short list of preconditions that generalize to any new geography. A firm needs legal permission to operate and a job to be done that exists there. It needs a solution adapted to local preference, which is nearly always required (even Coca-Cola adjusts its formula by region, and a twelve-pack of paper towels is not a product for a flat of forty-five square meters in Hong Kong). It needs enough autonomy in local decision making to move at local speed. And, decisive, it needs alpha assets that confer advantage against local competitors who have everything else: scale, a global brand, or a solution they cannot match. A newcomer without one of these has brought a knife to a gunfight.
The other directions follow the same logic. A firm can sell something more to the customers it already has, as Microsoft did when it attached Teams to Office. It can extend a brand into a category where the brand's promise still means something: Starbucks into packaged coffee and breakfast food, not into footwear. It can use its scale, as Interface, a maker of carpet tile for airports and offices, did when it launched a consumer brand, FLOR, that never mentions its parent and is made on the parent's lines at the parent's cost. A firm can also push a proven solution into a new job; Amazon's internal computing infrastructure became Amazon Web Services, the most valuable thing the company owns. Because a solution pushed into a job it was not designed for is often the wrong solution for that job, this direction takes care. Can the fifth flywheel point to a direction for growth? Sure. Toyota's production system is a culture of engaged people solving problems. Toyota now supplies that culture, through a dedicated unit, to organizations as far from automobiles as a children's hospital.
The direction to avoid is the one a firm's assets do not reach. Around 2000 the book retailer Borders, facing Amazon, had a respected brand, excellent locations, a strong balance sheet, and real capability in managing a temporary workforce and a high-variety inventory. None of those assets helped online. Borders handed its online store to Amazon to run, spent the decade in stores that mattered a little less each year, and filed for bankruptcy in 2011. Barnes and Noble, with similar assets, spent heavily on the Nook to fight the Kindle, where its assets did not reach, and found its durable business running college bookstores, where all of them did. The test I give executives is the one a venture investor would apply: knowing the firm's assets and not its history, would you fund its plan to enter this market? If the plan depends on assets the firm does not have, the answer is no, however attractive the market.
For a Person
The same tests apply to a career, with two differences. A person cannot spread the work across divisions or replace themselves, so the limits of a finite life bind harder. A person's objective is also not return on capital (chapter 9 takes up what it is), so for now I assume you want to do work you are good at and valued for.
Finding the setting begins with an inventory of capabilities, but most people cannot see their own combination from the inside. My own effort is vivid to me and the alternatives are not. Herminia Ibarra's research on career change, built on interviews with people who had made one, found that people rarely discover a better fit by analysis. They discover it by trying: taking on a project outside the current role, working with a new group, joining something on the side, and noticing where the work comes easily and is welcomed. These experiments are cheap relative to the commitment they inform. If you want to know whether your combination is rare in some other room, go and stand in the room.
Two more considerations follow from the mechanics of chapter 4. The first is that early choices compound: the setting a person enters first shapes the opportunities they are offered next, through every channel of cumulative advantage. So the first choice of setting has consequences out of proportion to the information available when it is made. The second is that switching has a cost that rises with the investment already made. A specialist who has spent a decade building an asset that is valuable in one setting and ordinary in another faces a real loss in moving. If the setting is wrong, though, the loss is not a reason to stay. It is a reason to decide early, and to run the cheap experiments before the investment is large.
The three questions that I ask a founder to answer in an advantage thesis in chapter 7 apply equally to a person's next five years. What is my advantage now, and who pays for it? What will it be in five years, given what I am building and what the setting will do to it? What must I do in the next two years to get from the first answer to the second? The questions are harder for a person, who has to answer them from inside their current setting. A person who cannot answer the second question is building an asset for a setting they have not chosen.
Comparative fit also applies among collaborators, inside a team as well as across settings. A founder who is a good engineer and a poor seller is better off with a seller than with a sales course. Together the two hold a combination that neither has alone. Most of the complements a person needs are other people.
The Person and the Model
What does the framework say about a frontier language model, a tool that in 2026 anyone can use? The model is performance enhancing and not hard to acquire, so it fails the second test of chapter 2, as electricity does. The economist Paul David made the historical point about electricity: having it conferred no advantage on a factory, and the gains went, decades later, to the factories redesigned around it and to the people who knew how. The returns to a universally available tool go to whoever holds the complements, and the question for a person is which complements those are.
The first of the four I would name is access to the customers, the problems, the decision rights and the data that make a model's output worth something. The second is accountability: someone must sign the work, and the signature carries a reputation that took years to build and that the model does not have. The third is judgment, which comes in two parts that are different assets. One is knowing what to ask, which requires enough domain knowledge to pose the problem; the other is knowing whether the answer is any good, which is taste. The fourth is the social capability to get an answer adopted by people who did not produce it. I suspect that when generation becomes nearly free, the bottleneck moves to selection, and the person who can select well captures the surplus. The model is available to everyone, so it captures none. Picture a person with unusual taste and an unusual ability to carry a room, paired with a model that generates without tiring. Each part can be found separately; the combination is rare. By the chapter 2 definition, the asset is the pairing.
The evidence so far is narrower than that claim. In the studies that exist, the model compresses the differences between workers within a task. In a customer-support operation studied by the economist Erik Brynjolfsson and his colleagues, productivity rose about fourteen percent on average, with most of the gain going to the least experienced agents and almost none to the best. In an experiment with consultants, the below-average performers gained far more than the above-average ones on tasks inside the model's competence, while on tasks outside it, consultants using the model did worse. Both studies show the model leveling execution within a task. Does the model also widen the gap at the top, by rewarding the judgment that lets a person exploit it? That is the hypothesis this section rests on, and so far it remains a hypothesis. Chapter 3 told the story of the one study that seemed to prove it.
For young people the framework gives an uncomfortable answer. The entry-level worker's historical alpha asset was a willingness to do codifiable work cheaply while accumulating the judgment that would later be the real asset. Now the model does the codifiable work. Early payroll evidence suggests that since 2022, in the occupations most exposed to the tools, employment of workers in their early twenties has fallen relative to older workers in the same occupations, while the older workers' employment has not. If that holds, the model is removing the apprenticeship that built the complement it makes valuable. Two things cut the other way. The first is that starting from nothing is cheaper: one person with taste and initiative can now build and ship what used to take a team. For people with the complements, the set of opportunities has widened, even as the entry-level jobs for people without them have thinned out. The second is that friction has fallen. Lower friction shortens the window in which a better product is an alpha asset and raises the value of the assets a model cannot build any faster: a customer base, a reputation, relationships. A young person who understands this will spend less of the window polishing the product and more of it acquiring customers.
What Travels
So the answer is the same for the firm and the person. Go where your assets are rare and necessary and where the complements you lack are supplied. Before you go, find out which of your assets travel, since the ones that do not belong to the setting and stay behind when you leave. The firm entering a new geography carries its brand and its scale and leaves its local relationships at home. The analyst changing banks carries their method and leaves the platform. The young engineer leaving a famous employer carries their judgment and leaves the name on the business card, and then discovers, as I did, how much of the inbound mail had been addressed to the name.
Notes
the New York Times reported from the Milan men's wear shows "Flying Shirttails: The New Pennants of Rebellion," New York Times, July 20, 2004. The 2022 draft quotes the passage on the untucked shirttails of the American men's wear press and buyers.
James March named the general tension exploration and exploitation James G. March, "Exploration and Exploitation in Organizational Learning," Organization Science 2, no. 1 (1991): 71–87.
Lele Sang and I studied eight multinational companies entering China Lele Sang and Karl T. Ulrich, Winning in China: 8 Stories of Success and Failure in the World's Largest Economy (Philadelphia: Wharton School Press, 2021). The preconditions for success in a new geography are adapted from that book.
Herminia Ibarra's research on career change Herminia Ibarra, Working Identity: Unconventional Strategies for Reinventing Your Career (Boston: Harvard Business School Press, 2003; updated edition 2023). Her findings rest on qualitative interviews with people who changed careers and emphasize experiment and new relationships over analysis.
Paul David made the historical point about electricity Paul A. David, "The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox," American Economic Review 80, no. 2 (1990): 355–361.
a customer-support operation studied by the economist Erik Brynjolfsson and his colleagues Erik Brynjolfsson, Danielle Li, and Lindsey R. Raymond, "Generative AI at Work," Quarterly Journal of Economics 140, no. 2 (2025): 889–942. Average productivity gain of about 14 percent, concentrated among less experienced and lower-skilled agents.