When Friction Collapses

product. We call it services. We build careers in it. We teach children to pursue it. Then we act shocked when the bill rises and the speed falls. The prevailing belief is that the service economy is the sign of a mature society. The factory fades, the field shrinks, the office grows, and the work becomes cleaner, safer, smarter. There is truth in that story. The trouble is that it hides a second truth. A large share of services is not care, expertise, hospitality, or creative craft. It is paid friction. It is paid delay. It is paid intermediation between people, systems, and decision rights. “What would have to be true for this outcome to keep repeating.” If you answer that honestly, you end up in a place most leaders do not want to go. You end up forced to admit that we built an economy that monetizes mistrust. We built a world where the cost of letting things happen without permission is so high that we pay people to stand between. We call it professionalism. We call it process. We call it being responsible. We also call it jobs. That is the part that is ending. Friction is paid delay, dressed as responsibility. Friction is not a moral failing. Friction is a technical and institutional condition. It shows up when evidence is scattered, when accountability is unclear, when liability is expensive, and when authority can only be exercised by a human signature. In that environment, the safest way to run an enterprise is to slow it down. You create gates. You create reviews. You create committees. You create roles whose output is permission. You move risk into procedure and then pretend it is under control. That machinery becomes a market. Entire industries form around it. Professional services are, at their best, a way to make complex coordination possible. They are also, at scale, a way to profit from the fact that the systems beneath them cannot be trusted to act. This is why “automation” is the wrong word for what is happening now. The threat is not that machines will do tasks. Machines have been doing tasks for a long time. The threat is that the economy’s largest paid layer is a layer of intermediation, and intermediation is what collapses when authority becomes enforceable in software. The moment permission is computable, the moment reasoning is auditable, the moment evidence can be bound to action, the economy stops paying people to be the glue. That is why the service economy, as we currently measure it, does not just get thinner. It changes shape. The parts that were real human service remain. The parts that were friction markets lose their pricing power. The headlines will call it job loss. The deeper event is a redefinition of value.

The world that follows is not post work. It is post proxy. The Labor Economy Was Always Here People hear labor economy and picture hard hats. That tells you how far we let language drift. Labor is not a class signal. Labor is the work that changes reality. Labor is the act that produces an outcome, physical or cognitive, local or global, human or machine assisted, but accountable in the record. The friction economy pays you for managing interfaces. The labor economy pays you for producing outcomes. That is not nostalgia. That is mechanics. In the friction economy you can be valuable while being several degrees away from the work. You can route decisions, coordinate calendars, reconcile systems, draft documents that protect people from blame, and make a career out of managing the distance between signal and action. In the labor economy distance stops paying. What pays is the ability to act, or the ability to make acting safe for others. That is why this change is not evenly distributed. It hits the center of white collar work first because the center is where permission lives. It also hits the entry level first because entry level has been, for decades, the training ground for intermediation. A country can survive a wave of automation that replaces repetitive tasks. It can even celebrate it. A country strains when automation replaces the on ramp into legitimacy. This is why the social risk is higher than the technology risk. A society can absorb the loss of certain tasks if it can still offer a path from youth to competence to stability. When the first rung breaks, resentment becomes rational. Authority is what lets work move without supervision. Authority is the unit that decides who can touch what and why. In most enterprises, authority is not encoded. It is inferred. It is negotiated. It is political. It lives in org charts, in tribal memory, in who knows who, and in the fear of being the person who signed the wrong thing. In that world, the safest strategy is to require more signatures. When you do that, you do not just slow the firm. You create a labor market. You create roles that exist because the system cannot be allowed to act. You hire humans to be circuit breakers. That circuit breaker class became modern services. It grew until it became the economy’s dominant share, and we told ourselves it meant we were advanced. Some services are advanced. Many are a tax.

The Service Economy Was the Friction Economy Wearing a Suit

A service can be real labor delivered as a service. Nursing. Teaching. Childcare. Emergency response. Repair. Craft. Food. Hospitality. Those are human outcomes. They are often hard, time bound, and consequence heavy. They do not vanish because software improves. But a service can also be coordination labor. The work whose output is alignment. The work that exists because systems do not agree, because evidence is not trusted, because policy is ambiguous, because liability is expensive, and because no one wants to be blamed. That layer is enormous. It shows up in law, accounting, compliance, claims, audit, procurement, HR operations, program management, and a large portion of what passes for corporate work. The test is simple. If the role exists mainly because the enterprise cannot let a process run without a human in the loop, it is friction. If the role exists mainly because evidence is fragmented across tools and teams, it is friction. If the role exists mainly because the organization has substituted ritual for encoded permission, it is friction. This is not an insult. It is an audit. Once you see it, you cannot unsee it. You notice how much of an operating review is not about operating, but about explaining. You notice how often a company pays its best people to sit in rooms and recite what everyone could already see if the systems were connected and trustworthy. You notice how often the same decision is made multiple times in multiple meetings, each time with a new cast, because authority is not bounded and evidence is not portable. That is the definition of friction. The second order effect is worse. When work cannot move, the organization learns to live with delay. It treats delay as prudence. It treats prudence as virtue. Then it treats the inability to move as normal. This is how firms die while feeling responsible. When permission is political, meetings replace execution. The Counterargument That Must Be Faced There is a strong counterargument to everything above. It is the argument that friction is not waste. It is safety. It is ethics. It is caution purchased with time. In many domains, that is true. Aviation is safe because it is slow where it must be slow. Nuclear is safe because it is governed by constraint. Medicine, at its best, is safe because it holds action behind evidence. Finance is

stable, when it is stable, because certain actions require sign off. In these domains, “move fast” is not courage. It is negligence. So the question is not whether friction should exist. The question is whether we can separate necessary constraint from accidental bureaucracy. The question is whether we can encode permission precisely enough that the system moves at speed inside safe boundaries, while remaining slow where slowness is the point. The friction economy does not make that distinction. It cannot. It uses human intermediation as the universal tool, which means the only way to buy safety is to buy delay. That leads to a tragic outcome. The same processes that protect in one context become a tax in every context. When leaders say, “We are highly regulated,” they are often telling the truth. They are also sometimes using regulation as an excuse to keep permission ambiguous, because ambiguity protects careers. A better world is one where constraint is explicit, and therefore enforceable. That world can move faster while also being safer, because it does not require humans to stand between every action and every consequence. This is where the definitional discipline matters. If a system can act without enforceable boundaries, it will create faster incidents. If a system can act within enforceable boundaries, it deletes intermediation without deleting safety. That is the difference between a tool and a hazard. If It Cannot Shape an Outcome, It Is Not an Agent The loose word in the market is agent. Everyone wants it. Everyone is selling it. Most of what is being sold is a chat interface over a pile of documents. An agent is not something that speaks. An agent is something that can shape an outcome. If it cannot shape an outcome, it is not an agent. If it cannot shape an outcome, it is not an agent. This definition sounds strict because it is strict. It is also necessary because the economic consequences are enormous. A system that only speaks does not collapse friction. It adds a new layer of interpretation. It may even create more meetings because now everyone argues about what the model “meant.” A system that can act changes the accounting. It deletes the queue. It removes the need for repeated human permission routing. It compresses the time between signal and intervention.

That compression is where the service economy begins to thin. Not because people stop needing service. Because the service that was paid delay stops being paid. Authority becomes the new currency. Not authority as ego. Authority as bounded permission tied to evidence and audited in the record. When that is present, work moves. When it is absent, work waits, and we hire people to manage the waiting. This is the mechanism that reshapes the world. It is testable. You can audit it in a company without asking anyone how they feel. Look at the workflow. Look at how many approvals are required. Look at how often exceptions trigger meetings. Look at how many roles exist whose primary artifact is a document that allows someone else to act. If that count is high, you are looking at a friction economy operating inside a firm. Here is the falsifiable prediction, and it should make any serious leader uneasy because it is not flattering. In the near term, the firms that adopt enforceable permission inside their core workflows will show a measurable drop in meeting load tied to routine decisions, and a measurable rise in direct throughput per professional headcount. If that does not occur, then the core claim in this piece is wrong, and the promise of agentic execution is mostly theater. That is a prediction worth being embarrassed by. It is also the credibility tax. The New Ladder Breaks at the Bottom Every society teaches a career story. For the past generation, the story went like this. Get educated. Get credentialed. Move into the service economy. Move away from physical labor. Move toward knowledge work. Climb in a firm. Become a manager. Become a leader. Enjoy stability. That story is breaking because much of what we called knowledge work was coordination work. It was work that existed because systems could not be trusted, and because permission was not enforceable. The first people to feel it are young. Not because they are less capable, but because they are the ones who traditionally entered through roles that were safe precisely because they were intermediation. The junior analyst who builds decks. The coordinator who schedules. The assistant who routes. The entry level marketer who drafts and edits. The junior developer who writes basic code. Those roles trained people to become the next layer of glue. If glue is being automated, the apprenticeship pipeline into the middle class changes shape. This is where leaders make an error. They tell young people to “learn AI.” That is not wrong, but it is not enough. Tools are not a career. Outcomes are a career. The question is what kind of outcome you can own, and whether you can prove it.

So the on ramp divides. One path returns to physical competence. Trades, maintenance, construction, energy systems, field service, advanced manufacturing. The work is local, the consequences are real, and the automation that arrives tends to complement rather than replace because the world is still made of atoms. Another path runs through care. Nursing. Teaching. Therapy. Elder care. Childcare. The work is emotionally and physically demanding. It will also be chronically needed in aging societies, and it becomes more valuable, not less, as friction is stripped away from everything else. A third path is hybrid technical work, but not in the romantic sense. Not “be a coder” as identity. Be a builder who can define constraints, validate outputs, and operate systems that act. When code generation becomes cheap, what becomes scarce is trustworthy deployment and accountable operation. A fourth path is direct entrepreneurship. Small teams that form around a problem, execute, and dissolve. This will grow because lower friction makes coordination cheaper. It will also be brutal because platforms concentrate distribution and push risk onto individuals. None of these paths are guaranteed prosperity. They are options. The societies that do well are the societies that build low shame pathways between them. The societies that do poorly are the societies that keep pretending the old ladder still reaches the roof. Mid-Career Is a Sorting Event Mid-career professionals often have a mortgage, children, obligations, and a calendar that cannot absorb a heroic reinvention. They are also the ones whose roles most often contain both real judgment and pure intermediation in the same job. That blend matters because it determines whether a role becomes more valuable or less valuable as friction collapses. If your day is spent making decisions that no one else can make because you hold domain knowledge, context, and accountability, you are close to labor. You are close to outcome. Agents will make you faster. They will remove drudgery. They will widen your span of control. If your day is spent brokering decisions that someone else could make if permission were explicit and evidence were portable, you are in friction. Agents will compress your role into a smaller number of higher authority operators. This is why “augmentation” can still be job loss. A role can be augmented and still require fewer people because the throughput per person rises. That is not evil. It is arithmetic. The moral question is what we do with the surplus, and whether we build transitions that respect human dignity. In the labor economy, prosperity across a lifetime stops being a one time credential event. It becomes a cycle. You build a base skill. You deepen domain ownership. You adopt tools that

increase output. You then move upstream into defining constraints, teaching others, and owning the consequences that the tools cannot carry. People who refuse that cycle will still find work. They may not find stability. The economy will pay them for what remains scarce. Scarcity will not be resume language. Scarcity will be accountable competence. Care Work Becomes the Benchmark, Not the Exception There is a deep irony in the next era. We treated care work as low status while building an economy that celebrated intermediation. The labor economy reverses the price signal. As friction markets thin, the relative scarcity of care and presence rises. A society that has an aging population cannot do without clinicians, caregivers, and teachers. That work is not infinitely compressible without losing what it is. You can remove paperwork. You can remove duplicate documentation. You can reduce wasted time. You can improve planning. You can personalize learning materials. None of that removes the human need. This is why the labor economy does not mean every job becomes physical. It means the jobs that remain valuable have a higher share of direct responsibility for human outcomes. That is the opposite of the friction economy, which pays handsomely for procedural distance. The trap is that care can still be crushed by bureaucracy. If we use AI to produce more forms, more compliance artifacts, more documentation, then we do not free clinicians and teachers. We trap them in a faster admin machine. This is how societies take a tool meant to reduce friction and turn it into a higher speed version of the same waste. This is why permission and authority matter even in human service. When a teacher cannot adapt a lesson without permission, the classroom becomes slow. When a nurse cannot execute a care plan without fighting a billing system, the hospital becomes slow. Friction steals the scarce resource, which is the human nervous system. The labor economy, done correctly, reprices care as essential and strips the admin load that has been treated as normal. Platforms, Concentration, and the Return of a Different Kind of Friction Lower friction does not guarantee decentralization. It can also produce concentration. When coordination becomes cheap, the value of controlling interfaces rises. Identity. Payment rails. Distribution. Compliance scaffolding. Data gravity. The pipes matter. The platform that controls the pipe can charge rent. So there is a second counterargument that deserves respect. The end of one friction economy can produce another. Instead of managers routing permission, platforms route access. Instead of internal bureaucracy, you get external gatekeepers.

This matters because it shapes whether the labor economy becomes a renaissance or a revolt. If surplus created by friction collapse is captured by a thin ownership layer, social legitimacy collapses. People will not accept a world where their intermediation job vanished and their replacement is a low paid gig managed by an algorithm they cannot appeal. This is why the question of prosperity is inseparable from the question of institutions. It is not enough that work exists. People need a credible path. They need predictable rules. They need training systems that do not bankrupt them. They need portable benefits. They need a social contract that treats labor as dignified. If we do not build that, fear becomes organized. In a world where trust is already thin, organized fear is gasoline. The Two Boardroom Diagnostics That Tell You Which Economy You Are In Every executive team claims they want speed. Most of them are running a machine designed for permission theater. The difference is visible if you ask the right questions, in the right order, without letting the answers become slogans. When an exception hits, who can act without convening a meeting, and what evidence must exist in the record for that action to be legitimate? When a decision is delayed, is it delayed because the evidence is missing, or because authority is ambiguous, or because nobody wants to sign their name to the consequence? Those questions are not philosophical. They are operational. A firm that cannot answer them precisely will pay for that ambiguity in time, in margin, and in morale. Here is the second diagnostic, and it is more personally threatening, which is why it matters. If you removed every role whose output is routing, coordination, reconciliation, and defensibility, which outcomes would actually fail, and which would simply happen faster? Most leaders have never been forced to answer that honestly because the economy rewarded them for hiring more glue. In the labor economy, the audit arrives anyway. The only question is whether you conduct it with discipline or whether the market conducts it for you. Universal Skills, After the Proxy Layer Thins The old model of skill was accumulation. Accumulate knowledge. Accumulate credentials. Accumulate titles. That model assumed the market would pay you for what you know, even if you are far from the outcome. The labor economy pays you for what you can do, and for what you can be trusted to own. That sounds like a motivational line until you translate it into what actually becomes scarce.

Judgment under constraint becomes scarce. Not brilliance. Responsibility. The ability to decide inside boundaries and explain why the decision was legitimate. Systems thinking becomes scarce. Not the word. The practice of seeing how incentive, delay, and feedback loops produce results, and of designing work so that the loop closes fast enough to matter. Human trust building becomes scarce. Not charisma. The ability to build durable cooperation under time pressure, conflict, and uncertainty. Craft becomes scarce. The ability to produce quality repeatedly, whether that quality is a weld, a lesson, a diagnosis, a procurement decision, a safety intervention, a field repair, or a piece of writing that holds up under scrutiny. Agency becomes scarce. The ability to keep learning, keep adapting, keep taking ownership when the environment changes and the old path stops paying. These skills are not new. That is the point. The past is not the enemy. The enemy is pretending yesterday’s proxy signals will protect you tomorrow. Education Must Prove, Not Promise Education systems were built for a world where information was scarce, and where the employer absorbed the cost of turning a graduate into a capable worker. That world is fading. When the entry ladder compresses, education faces a verdict it has avoided for decades. It must become accountable for competence, not attendance. It must teach people how to produce outcomes, not how to survive school. This does not mean everyone needs a trade. It means every path needs proof. A portfolio. A demonstrated skill. A real artifact. A practicum that has consequences. A supervised environment where mistakes teach without destroying lives. It also means education cannot stop at early adulthood. In a labor economy, the person who prospers is not the person who guessed correctly at eighteen. It is the person who can retool, more than once, without losing dignity or financial stability. A society that treats adult learning as a private luxury will widen inequality. The people with capital will retool. The people without capital will be told to retool and then punished for not doing it. That is how a nation fractures while insisting it values opportunity. So education becomes a lifelong institution in practice, not as rhetoric. It becomes modular. It becomes tied to work. It becomes less about sorting and more about building capability. That change is not a progressive dream. It is economic necessity. The Social Consequences Are Allocation, Not Innovation

The largest social fights of the next era will not be about whether the technology works. The technology will work well enough to delete intermediation. The fights will be about who captures the surplus and who bears the transition cost. If friction collapses, throughput rises. Costs fall. Decision cycles compress. That creates surplus. The question is where the surplus goes. A healthy society uses surplus to raise wages for direct labor, lower prices for essentials, invest in care and infrastructure, and fund pathways for people to move from proxy work into outcome work. An unhealthy society concentrates surplus into ownership layers, builds platforms that charge rent on access, and tells displaced workers to learn new skills while offering no credible bridge. That society will not remain calm. It will not remain stable. It will not remain governable without coercion, and coercion is just friction by another name. This is why the labor economy is not merely economic. It is civic. It changes what we respect. It changes what we pay. It changes what we teach. It changes how we measure status. It changes how we define a good life. A service economy that was really a friction economy produced a strange moral outcome. It taught people that dignity comes from distance. Distance from physical work. Distance from consequence. Distance from risk. Distance from responsibility. The labor economy reverses that. It returns dignity to the person who can act, fix, build, care, and own. That reversal will feel like liberation to some. It will feel like humiliation to others. If we treat the second group with contempt, we will deserve the instability that follows. The Ending That Leaders Avoid Most executives want to close pieces like this with actions, steps, playbooks. That impulse is part of the problem. It is the same impulse that turned governance into ritual. It is the desire to turn fear into procedure so you can feel safe. The real ending is simpler, and harsher. Friction is an economic condition. It can be engineered down when permission is enforceable and authority is explicit. When that happens, intermediation stops being priced as value. The service economy thins where it was paid delay. The labor economy becomes visible again, and it becomes the basis for prosperity. You can debate whether you like that world. You cannot veto it.

The only open question is whether we build the institutions that let humans prosper inside it, or whether we let the market do the sorting in public. The bill is already in the room. References This essay draws on Douglass North’s Nobel Prize lecture in 1993 for the clean definition of transaction costs as the costs of specifying exchange and enforcing agreements, because friction is not a metaphor here, it is an economic mechanism with measurable consequences. It uses Ronald Coase’s 1937 argument on why firms exist when coordination is costly, because collapsing coordination cost changes firm boundaries and alters the labor market for intermediation. It leans on William Baumol’s work on cost disease, because care, teaching, and other time bound human services become more central once bureaucracy is stripped away, and their economics become harder to ignore. It draws on the IMF’s January 2024 analysis of AI’s exposure across global employment and the distributional risk that follows, because the central issue is not capability but who captures surplus and who bears transition costs. It uses the ILO’s 2023 global analysis and the 2025 refined index on generative AI exposure, because clerical and coordination heavy work is unevenly exposed and the gendered distribution of those roles has direct social consequence. It relies on the World Economic Forum’s Future of Jobs 2025 reporting on labor market churn through the end of the decade, because net job creation can coexist with violent personal discontinuity when the entry ladder compresses. It anchors the education argument in UNESCO’s call for a renewed social contract for education and the OECD Learning Compass emphasis on agency and lifelong capability, because the prosperity problem is institutional, not motivational. It also uses OECD work on social and emotional skills across adulthood, because durable prosperity in a low friction economy depends on judgment, cooperation, and accountability as much as technical fluency. Foundationally, it is built on Michael Carroll’s own operating doctrine and definitional discipline on agents, authority, permission, and outcome shaping, because the argument depends on the claim that authority can become enforceable in software and that this is the mechanism that collapses paid intermediation.