What Agency Is

The chamber of crafts is in a gray building with a good coffee machine and a numbering system that has never once called Abdellatif’s number early. He has taken a half day of vacation for this — his third half day for this file — and he sits with a plastic folder on his knees containing, in the order the checklist demands: his Moroccan electrician’s diploma, a certified translation of his Moroccan electrician’s diploma, an apostille, two employer letters from Casablanca, and a printout of an email in which a caseworker confirmed, four months ago, that the file was complete.

The woman at counter 6 is not unfriendly. That is somehow the hardest part.

“The assessment is still with the technical examiner,” she says. “There’s a backlog.”

“In February you said eight to ten weeks.”

“That was the estimate in February.” She looks at the screen, and something in her face suggests she genuinely wishes the screen said something else. “You could be invited to a skills analysis instead — a practical assessment. But that has its own waiting list.”

Abdellatif came to Germany from Morocco with a trade in his hands. He has rewired farmhouses outside Casablanca during power cuts, holding a flashlight in his teeth. He has passed, for practice, three online sample tests for the German journeyman’s exam — in German, in the evenings, after shifts at the logistics center where he scans and loads other people’s appliances. He knows three-phase current does not care which country certified you.

“Is there anything I can do to move it?” he asks. “Anything at all. I’ll take the assessment tomorrow. Today.”

“I understand,” she says, and he believes her, and it changes nothing. “You’ll receive written notice.”

Outside, he stands in the parking lot and does what he always does: converts the anger into arithmetic. Every month the file sits, he earns a warehouse wage instead of an electrician’s wage — call the difference €900, after taxes. Every month is also a month he cannot honestly answer the question his girlfriend Lea has stopped asking out loud, and a month his mother in Casablanca hears from his sisters that he is “still carrying boxes.”

He is not lacking skill. He is not lacking drive — ask anyone who has watched him work a double shift and then sit with a grammar app on the night bus. What he lacks is stamped on no checklist: the room to use what he has.

A definition worth the weight

This essay builds the model the whole series runs on, so it is worth being precise at the start.

Agency is the socially produced capacity to direct your own course: to set goals, act on them, persist through difficulty, and adapt when things don’t work.

Every word is load-bearing. Capacity, because agency is something you can have more or less of, and can build — not a binary. Direct your own course, because the object is your life’s heading, not your mood or your mindset. And socially produced, because this is where popular talk about self-direction goes wrong. Agency is not a personality trait. It is produced by upbringing, schooling, institutions, and practice — which means it can be built and eroded, and much of both happens through no decision of your own. In Education is Broken I called agency the load-bearing wall of any educational model: if it is missing, everything on top collapses. This book asks the same wall to hold up a whole adult life.

The definition has a respectable lineage, and each source contributes a piece the others lack. Albert Bandura, who did more than anyone to make human agency a scientific subject, described agents as people who make things happen rather than merely undergo them — and found that self-efficacy, the belief that you can execute what a situation demands, is built primarily through mastery experiences: doing difficult things and watching them work. Not affirmation; evidence you give yourself. Amartya Sen’s capability approach contributes the distinction that keeps this book honest: what matters is not what a person can nominally do but their real freedom to do it — the gap between having a skill and having the opportunity to use it. Abdellatif in the parking lot is a walking illustration. And Edward Deci and Richard Ryan, in self-determination theory, mapped the motivational core — autonomy, competence, relatedness — and explained something the other two don’t: why a person can have skill and opportunity and still not move.

Three traditions, developed independently, pointing at three faces of one object: the three dimensions of agency.

The three dimensions

Capability is the ability to function in life, professionally and privately — and to keep functioning as conditions change. The master capability is self-directed learning — figuring out what you need to know, acquiring it, applying it, with no institution assigning it — because it renews all the others. Every specific skill is depreciating; the capacity to build new ones is the only asset that appreciates with use. Barry Zimmerman’s research on self-regulated learning shows this capacity has learnable components and develops through cycles of challenge, feedback, and reflection, not instruction alone. Three others share the dimension: working with intelligent machines — knowing what to delegate to AI and what never to; judgment — deciding well where evidence is incomplete, with Kahneman and Tversky’s work as a map of the ways it systematically fails; and sensemaking — distinguishing signal from noise, evidence from manipulation. What is not in this dimension matters as much: your stock of domain knowledge. It is inventory, not capability. Inventory expires.

Room to act is the objective space of action you have — Sen’s real freedom, made concrete in three arenas. Economic room: income, buffers, obligations, condensed into one navigator’s number this book calls runway — the time your resources buy at your current burn rate. Two people with identical skills and courage, one with eighteen months of runway and one with eighteen days, cannot take the same actions. Political and institutional room: rights, legal status, credentials, and the gatekeepers who administer them — and, just as decisive, contestability: whether you can appeal, escalate, or route around a decision made about you. Albert Hirschman’s Exit, Voice, and Loyalty is the lens: you can leave, speak up, or stay quiet — and real power lies in having exit available, because voice is only taken seriously from people who can leave. And social room: networks, family systems, communities — the people who can carry you, vouch for you, warn you, and whose expectations also bind you. Mark Granovetter’s finding on weak ties belongs here: opportunities disproportionately arrive through acquaintances, who live in different information worlds than your close friends. Social room is the strangest arena because it is resource and constraint at once — Abdellatif’s family is his safety net and his heaviest obligation, in the same breath.

Room is unevenly distributed and only partly self-made. You did not choose your passport, your parents’ savings, or your name’s effect on a landlord. But room can be read — most people are wrong about how much they actually have, in both directions — defended against quiet erosion, and expanded, deliberately, at the margins, over time.

Will is the disposition to act: ownership of your own course, tolerance for uncertainty, and the readiness to trade the illusion of security for real options. Will is the dimension polite analysis prefers to skip, because it sounds like character talk. It isn’t. Deci and Ryan make it tractable: the will to act is fed by autonomy (this is my course, not one assigned to me), competence (I have evidence I can do hard things — Bandura again), and relatedness (people who matter to me are with me in this). Starve any of the three and will decays — predictably, in anyone. Scarcity attacks from another side — Mullainathan and Shafir showed it consumes the very cognitive bandwidth self-direction runs on. Will’s positive form is what this book calls the adventure stance: life as an ongoing expedition whose route is discovered en route, not a linear progression to be completed. Planning still happens — but as preparation, not prediction. And will is not intensity — not loudness, not risk appetite, not hours worked. A quiet person methodically expanding her options has more will than a manic job-hopper fleeing forward.

The interaction rule

Here is the sentence that makes the model a model rather than a list: agency is limited by its weakest dimension.

The three dimensions do not add; they gate. Each failure pattern has a name and a face. Capability without room is frustration — Abdellatif, folder on his knees at counter 6. The skill is real; the door is closed. People in this pattern get told to upskill, which is like telling a docked ship to improve its engine. Room without will is drift — Sonja: mid-thirties, insurance economics degree, ten years of doing everything right at a large insurer whose restructuring memos have started deleting the career ladder above her. If she does nothing — savings, strong credential, intact network — every month the default of staying put quietly makes next month’s staying more likely. Drift is agony in slow motion precisely because nothing is visibly wrong. And will without capability is recklessness — the person who storms out of the corporate cocoon, the shell that feels secure while it narrows options, with no runway calculation, no rebuilt skills, and a plan made of adrenaline. Some survive it. The survivors write books; the base rate does not.

The rule dictates strategy: work on your weakest dimension first. This sounds obvious and is systematically violated, because people prefer to invest where they are already strong — the capable take another course, the resourced buy another buffer, the willful make another leap. The rule also explains what Education is Broken called the agency divide: when opportunity is distributed without regard to all three dimensions, it flows to those who can already convert it. Two people, same AI tools, same open web, same nominal access — opposite outcomes. Access was never the binding constraint; the weakest dimension was.

Run both people through the model and the reason this series follows them becomes visible. Sonja has room — real runway (never yet calculated), full institutional standing, a well-placed network — but aging capability and a hidden will deficit: eleven years of being chosen, for the program, by the mentor, onto the list, have quietly transferred ownership of her course to the choosers. Abdellatif has will — he crossed continents on it — and high practical capability, including the navigational intelligence of someone who has never been handed a map; what he lacks is room, and least of his own making. Her weakest dimension is invisible from outside; his is the only thing visible. Neither is a victim; neither is a hero. Most readers are somewhere on the line between them.

What agency is not

The word “agency” has been colonized by people selling things, so the model needs its borders marked.

Agency is not hustle culture. The grind aesthetic — rise at five, monetize everything, sleep when you’re dead — is mostly will performed as theater, and it destroys the other two dimensions: it burns capability (no time to learn deeply) and it burns room (health and relationships are room). Sustainable agency looks less like sprinting and more like seamanship: unglamorous, rhythmic, attentive.

Agency is not positive thinking. No belief, however sincere, changes a visa status or a bank balance. Bandura’s self-efficacy is sometimes misquoted as “believe in yourself”; his actual finding is nearly the opposite — durable confidence is downstream of evidence. Optimism without mastery experiences is not agency; it is exposure.

And agency is not a license to blame. This is the border that matters most. If agency decides outcomes, it is tempting to run the logic backward: bad outcome, therefore insufficient agency, therefore your fault. The model rejects that inference — not as a courtesy, but by its own logic. Agency is socially produced: the childhood that builds self-efficacy, the schooling that teaches self-direction, the buffer that permits risk are unevenly given long before anyone chooses anything. And agency is gated by room: no disposition compensates for a revoked status or an empty account. When a systematically constrained person fails to flourish, the model’s verdict is not “she lacked will.” It is: her weakest dimension was chosen for her. An essay for individuals can say, honestly, both things at once — the divide is unjust and your position in it is partly workable. It must never collapse the second into a denial of the first.

The best square on the board

The internal mobility portal highlights new postings in green. Sonja has had one open in a browser tab since Monday: Project Lead, Claims Transformation — AI Deployment. 24 months, reporting to the COO office. She meets every listed requirement. She has read the posting perhaps ten times; she could recite it.

It is, she knows with professional certainty, the single best square on the board. Whoever runs the deployment learns the machines from the inside, meets the executive floor weekly, and becomes hard to cut — you do not, she reasons, lay off the person who holds the map of the new territory. She would advise anyone to take it. She has, in fact: this morning a colleague asked whether she should apply, and Sonja said obviously, immediately, and meant it.

Her own cursor hovers over “Apply.”

If she applies and is chosen, the old arithmetic resumes — chosen again, safe again, someone’s list again. If she applies and is not chosen — and the thought arrives with a physical drop, somewhere below the sternum — then she is officially what the restructuring memo implied: legacy. Better, whispers something old and fluent in her, not to submit the question than to receive that answer.

She notices the whisper. That is new — five weeks ago she would not have heard it as a voice at all, only as a reasonable-sounding decision to “wait for more clarity.” She sits with it for a moment, hand still on the mouse.

Then she closes the tab. Not the portal — the tab. There is a difference, she tells herself.

Room to act: maximal. Capability: sufficient. The posting stays open until the 15th.

---

Sonja and Abdellatif are constructed figures — personas built to carry real mechanisms, not case studies of real persons. The book’s method note explains how they’re used.