What remains of a craft

In 2010, I am twenty-five and I have just finished the year at more than 120% of my targets. I am still in the state of mind a business school installs in those who succeed there: the certainty that a career is a staircase, and that I have just climbed a step. So I go to see my director and announce that I want to manage. I insist. I claim to know exactly what needs to be done. He eventually gives in, in his own way: he grants me the right to recruit one salesperson, a single one, and to manage him.

For six months, I make every possible mistake. I take calls in his place. My feedback has neither rhythm nor form. His KPIs are vague, and he has no tool to progress with. Above all, without realizing it, I try to shape him into my own likeness. I pass on my phrases, my reflexes, my angles, as if I were talking to a kind of alter ego. I am not managing a person. I am duplicating a profile.

If my director let me have those six months, it was not only patience. It was that my individual results kept holding, and they masked the problem while deepening it. The more I performed, the more visible the gap between my trajectory and his became: he was not progressing with the same linearity, and I read that divergence as his failure, never as a failure of my management. The clone did not resemble the model, so the clone had to be defective. I was my only point of comparison. That, precisely, was the problem.

After six months, I go back to my director and admit I do not know how to do this. We plan training sessions, I start seeking advice, watching those who know how to work. I learned this craft in pain, and that pain was my first real teacher.

It took me a long time to see what that scene really contained. Two things, and they frame this triptych as it closes. The first is my mistake itself: believing that transmitting means duplicating yourself. That mistake has not aged a day. The second is a condition so ordinary at the time that nobody noticed it: time. My director gave me six months to be wrong, because the craft was stable enough for six months of mistakes to be an investment. That condition is what has disappeared.

A craft, still

The first paper of this triptych laid out the chronology: four regimes in six years, a craft that changed in nature. The second went down into the day-to-day and ended on an observation: the fundamentals of management were not replaced by AI, they were sorted by it. What remains is the question that sorting leaves open, and it is the most uncomfortable one. What is a craft, still, when every year expires part of what you have learned in it?

A craft, as it has always been understood, is a body of proven moves, and a community that passes them on with enough duration for the transmission to mean something. The journeyman learned moves that would hold for his whole life. So did the sales manager trained in 1995. It is the second half of that definition the period has destroyed: the duration. The moves I could show a beginning manager today will be sorted in their turn, and we both know it at the very moment I show them.

So either the craft disappears with the stability of its moves, or it was something other than its moves, and the current sorting is revealing what.

I lean toward the second reading, and I have a simple argument: my pipeline review. In 2012, it is a Monday-morning meeting, in a room, around a whiteboard. In 2021, it is a video call where I relearn how to read a team without seeing it. In 2024, it arrives pre-analyzed by an LLM and my work begins where its synthesis stops. In 2026, it has a little sister: the weekly review of agent outputs, scheduled, non-negotiable. Four forms in fifteen years. The move has changed four times. What it serves has never moved: making drift visible before it costs.

That is true of everything that held across the four regimes. The form expired at each rupture. The criterion that justified it came through. A craft is not the sum of its moves, which are the dated expression of something harder. That something is criteria of judgment: what counts, what drifts, what gets controlled, what gets left alone. Moves wear out. Criteria recognize one another from one regime to the next.

And a criterion is only recognized at the moment it costs. I walked away from a strategic account whose budget would have made the quarter, because the relationship was settling into a power struggle rather than a joint project, and what we would have signed was not what we claimed to sell. Refusing a number you have announced is pleasant for no one, neither in front of a team that had worked it, nor in front of a leadership that had counted it. That is exactly why it is a criterion and not a preference. A method is judged by what it produces; a criterion is judged by what you accept to lose in order to hold it. The moves of that negotiation, for their part, have since been sorted twice.

*A craft is no longer recognized by its moves. It is recognized by its criteria of judgment.*

What you transmit to the one who is starting out

Let us come back to my 2010 scene, because here it takes on a relief I could not see at the time.

My duplication mistake was recoverable for a reason that had nothing to do with me: even clumsily transmitted, my moves had value. They would remain valid for ten years. A salesperson trained in my likeness in 2010 was badly managed, but not badly equipped. Today, the same mistake produces damage of another magnitude. Transmitting your moves to a beginning manager means handing him an asset that depreciates faster than he can amortize it. The sequences I master, the agent settings I have calibrated, the stacks I know how to run: all of it will be sorted before he has finished learning it.

What we owe the one who is starting out is no longer moves. It is the criteria that will let him sort his own.

And among those criteria, there is one I place above the others, because it precedes every stack and will survive every rupture.

Every manager has at some point met the cascade trainers have taught for decades. There is what I mean to say, what I say, what the other hears, what he understands, what he retains, and what he does with it. Six states of the same message, and a loss at every step. It is textbook obvious, and yet I have never seen a manager hold it for more than a few weeks in actual practice, myself included.

An exchange under the second panel made me see that this cascade was only a special case. That paper described the simultaneity of registers: five languages, five rhythms, five criteria of evaluation in a single day. A reader added the missing piece. The loss does not bear only on messages. It bears on entire situations.

*The client, the collaborator, the leadership do not just live different times. They live different realities of the same moment.*

The same deal, at the same instant, is a strategic opportunity for the leadership, a negotiation bogging down for the salesperson carrying it, a reliability test for the client, and a forecast line for the agent system scoring it. None of these realities is false. None is sufficient. The manager's craft consists precisely in holding these realities together and turning them into a decision. That is where the clarity and lucidity expected of him reside, and they appear in no job description. Clarity is not a talent for expression. It is the result of work: having crossed the divergent perceptions of the same situation before deciding.

AI agents have added to this picture a category the classic cascade did not foresee. An agent loses nothing between what I say and what it hears: it executes exactly what is written, without filter, without mood, without charitable interpretation. All its loss is displaced upstream, into what I have not written. Where an experienced salesperson fills my silences with context he has lived, the agent fills its own with what it deduces. It is the same missing perception, and it produces opposite damage: the human understands less than what I said, the agent does more than what I wrote.

And this is where my 2010 scene delivers its real content. My duplication mistake was a mistake of perception. I saw only one reality of the craft, mine, with my linearity as the only yardstick. The salesperson I was managing lived another reality of the same open space, the same targets, the same weeks. I was incapable of perceiving it, and therefore incapable of transmitting anything that could reach it. Transmitting while understanding the stakeholders' point of view is not a supplement of kindness. It is the technical condition of transmission. You transmit nothing to someone whose reality you do not perceive.

The registers of the second panel thus cross the perceptions of this one, and the crossing draws the invariant core of the craft: multiple temporalities, lived as different realities, to be converged into decisions. The entities producing these perceptions change, and some are now artificial. The convergence, though, remains the work. A manager who masters it will cross the coming regimes. A manager who does not will perform until the next rupture, not beyond.

The second panel detailed two other pieces of the legacy, and I will not redeploy them here: the framing of autonomy task by task, which holds for a junior salesperson as for an AI agent, and the relationship to time, which decides what becomes of the hours the agents free up.

That single salesperson of 2010 was followed by many others. In fifteen years, I have managed dozens of salespeople, up to fifteen per team, each on his vertical, then managers who managed in their turn. What changes with scale is not the nature of the problem, it is its speed of propagation. A criterion badly transmitted to a salesperson produces a badly equipped salesperson. The same criterion badly transmitted to a manager produces an entire line applying the wrong rule, and it takes two quarters to notice. That is why I moved from transmission by example to transmission by written criteria. Example does not make it past the second floor.

But transmission has a second half, and it is the one my scene lights up in negative. My director taught me almost nothing directly. He gave me time, and he let the pain do the work. That was the pedagogy of a stable craft: six months of free mistakes, then lucidity, then training. A director in 2026 can no longer offer that, not out of harshness, but because the material changes faster than pain teaches. Six months of mistakes on an agent system are no longer six months of learning; they are three drifts settled in and a quarter of scrambled signal. Error remains the best teacher in the craft. It simply has to stop being a luxury granted by stability and become a constructed frame: perimeters where being wrong costs little, errors documented instead of digested in silence, reviews that turn failure into criterion. The second panel named this discipline for oneself. Transmitting means installing it in someone else from day one, instead of hoping it emerges from six months of pain you can no longer pay for him.

*We no longer transmit moves. We transmit what sorts them.*

What a career is, now

There remains the third question, the one that would be asked by the twenty-five-year-old salesperson coming to see me today as I went to see my director. What does a career look like in a discipline that keeps redrawing itself, and is it still worth entering?

The career business school promised me was a staircase: every skill acquired stayed acquired, the stock grew, the stock was the value. That is over, and not only for sales managers. A manager who started in 2020 has already seen three regimes. His stock of skills contains layers still warm and already expired. If he measures his value by what he knows how to do, he is getting poorer as he works.

What constitutes a career today is no longer the stock. It is a stranger asset, one that appears on no standard CV: the proof, repeated at each rupture, that you know how to relearn. The 2026 manager who commands a premium is not the one with twenty years of experience. It is the one whose twenty years have produced criteria, and who has demonstrated four times in a row that his criteria survived the sorting of his moves. Experience no longer holds value by accumulation. It holds value by what it has left behind that is transversal.

That asset can be detected, and I can say what I recognize it by since one precise recruitment. I was looking for a salesperson and receiving dozens of applications, many of them excellent on paper: the recognized schools, the high-profile startups, the methodologies mastered down to the fingertips. Reading them at length, I saw what they had in common. They were smoothed. The same sequences, the same frameworks, the same answers to the same questions, learned in the same environments. An impeccable stock, and an interchangeable one.

I kept an atypical profile. A person based in Lebanon, mostly anglophone, francophone all the same, with little experience in sales. Nothing of the stock the others presented. But in the interview, a will to learn and to succeed so dense it became contagious. I came out of the conversation wanting to work, which is not a frequent signal when you are the one recruiting. I bet on that energy against the logos. He did not just become good. He ended up among the best salespeople, to the point of being courted by my anglophone counterparts.

What that recruitment taught me was not a lesson in open-mindedness. It was a lesson in valuation. What I had bought that day was not a skill, it was a learning speed, and that speed turned out to be the only asset that compounded instead of depreciating. The impeccable stocks I had set aside were worth a great deal at the instant of the interview and a little less every quarter. Since then, that is what I look for first: less what a candidate knows how to do than the proof, somewhere in his path, that he has already learned something difficult starting from nothing.

And I watch for one precise reaction, the one no interview coaching prepares: what a candidate does when I take him onto ground he does not master. The one protecting his stock dodges and steers the conversation back to what he knows. The one who will survive the ruptures dives in.

So, to that young salesperson, I would answer yes. But not for the reasons I was given. Not for the staircase, which no longer exists. Not for the status, which protects from nothing. Yes, because this craft has become one of the most intense posts for the exercise of judgment there is: coordinating human and artificial intelligences, arbitrating simultaneously between registers that do not speak the same language, continuously sorting what still holds value. The craft he would enter is harder than the one I entered. It is also more interesting, precisely for the reason that makes it harder.

And I would tell him one last thing, which my director had not needed to tell me. Nobody will grant him my six months. He should not ask for them. He should demand instead what replaces them: a frame where error teaches fast, a manager who transmits criteria rather than recipes, and the right to watch his own moves expire without concluding that he has failed.

Inhabiting your ruptures, concretely

It remains to say what this definition changes in an actual week, because a craft is not practiced with formulas. Here is what the discipline this triptych has tried to name looks like in mine. Nothing that follows is a method. These are practices, built by successive corrections, and some of them will be sorted in their turn.

Relearning has a budget, and that budget is defended like a client meeting. Every week, blocked hours, outside production, to test what has just changed: an updated model, a new agent capability, a connector that changes what my orchestrator can read. I test on real cases, never in production flows. Part of the time the agents free up is reinvested in this test bench, and it is the only use of that time that repays its own debt: the agents free hours for me, and those hours are spent staying capable of managing them.

No trust is renewed tacitly from one version to the next. The second panel noted it: an agent's task-relevant maturity can regress from one update to the next. I have turned that into a mechanical rule. Every significant model update triggers a re-qualification of the agents concerned on their sensitive tasks, the way you would re-certify a collaborator coming back from a long absence. Most of the time, everything holds. Sometimes, a setting that had worked for months breaks silently, and the test bench sees it before the client does.

Errors are documented, mine included. Every corrected agent drift becomes a written rule, dated, versioned in the orchestration document. Every management error I own up to follows the same path. And twice a year, I reread my own rules from six months earlier. It is the most destabilizing exercise on the list: part of them is already obsolete, and that is precisely what you have to go look at. A manager who never rereads his rules still believes his craft is stable.

Crossing the perceptions has become an explicit move, no longer just an intuition. Before any decision that commits, I set down the realities in play: what this moment is for the salesperson living it, for the client, for the leadership, and for the agent system measuring it. A few lines, not a report. When I cannot fill in one of those lines, it means I am about to decide on a single perception. 2010 taught me where that leads.

And transmission begins on day one, with what I did not have and no one can offer anymore. Not six months of free mistakes: a perimeter where error costs little and teaches fast, mistakes put in writing instead of digested in silence, reviews where we sort together what has just expired. With the right frame, a junior learns in a few weeks what pain taught me in six months. This became a management obsession well before the first agent: condensing experience. I want you to learn in a few weeks what took me five years to understand, so that your own experience gets built on subjects I have not explored. Behind this obsession lies a conviction that is the exact inverse of my 2010 mistake: managing means making the people you manage become better than you, compartment by compartment of the game. And, as a conductor, not hesitating to recruit those you can already see are better on certain aspects of the craft, because that is precisely what makes a team perform beyond its manager's ceiling. In 2010, I was looking for someone who resembled me. The craft taught me to look for people who surpass me.

There is, finally, an exercise companies rush through, the values review, and I ended up bringing it back to where it produces something: into the relationship between a manager and each of the people he manages. Not the values displayed in the lobby. The ones that get verified in the work: the relationship to time, discipline, the will to progress, alignment with results. We revisit them regularly, explicitly, the way you review a pipe. When that symbiosis exists, everything else becomes adjustable, starting with the stack. That is the point the tool debates forget: the stack, agents included, is not the craft. It is an instrument of accomplishment in the service of what the company and the people who make it have decided together. A team aligned on its values adapts its stack in a few weeks. A team that is not can have the best stack on the market: it will execute, fast, something nobody carries.

None of this is spectacular, and that is the point. Inhabiting your ruptures does not look like a transformation. It looks like a hygiene: defended hours, re-tested trusts, reread rules, crossed perceptions, written errors, revisited values. No organization will ask a manager for this hygiene, because the recruitment grids, the annual reviews and the career plans were all built for the old craft and keep evaluating a stock. Nobody will be sanctioned for not keeping it. That is precisely what makes it a discipline rather than a procedure: it cannot be controlled from the outside; you impose it on yourself, or it does not exist.

One certainty remains, and it is the only one I keep from eighteen years in this craft. The first paper observed that you no longer learn a manager's craft, you learn to inhabit your ruptures. At the end of this journey, the phrase has stopped being an image.

*Inhabiting your ruptures was not a metaphor. It has become the definition of the craft.*

Article 3 · Sales management triptych · July 2026