Nadia had spent three days on the job description.

That was not negligible. She had run the group's human resources for four years and knew from experience that a poorly written job description cost six weeks of useless pre-selection, two offended candidates and a revision meeting nobody wanted to hold. So she had been careful.

She had printed the document in two copies. One for Isabelle. One for Laurence, whom the managing director had asked to include — for reasons Nadia had stopped trying to deduce in advance.

Isabelle read the first page without saying anything.

Laurence read the second with the attention of someone checking a technical text.

— Business lawyer, said Isabelle. Five to seven years' experience. Good knowledge of regulatory environments. Ability to work cross-functionally with operations.

— That's correct, said Nadia.

— Compliance certification appreciated.

— That's the standard profile in this market. Very legible for search firms. We generally receive good candidates with this type of wording.

— I don't doubt it. That's precisely the problem.

Nadia took note. Not because she understood, but because taking notes in difficult silences was a professional habit that had served her well.

— You want someone more senior. We can revisit the range.

— It's not a question of seniority.

— Sector, then. We can target profiles from regulated industries.

— Neither.

Nadia set down her pen.

— Then tell me what you're looking for.

Isabelle hesitated. Not from imprecision — she knew what she had been feeling for several months. But she had not yet managed to formulate it without it sounding like a criticism directed at someone in the room.

— Someone whose first reflex is not to tell us what could go wrong. Someone who starts from where we want to go and looks for how to get us there despite the constraints. Someone capable of producing an option and defending it — not a list of reasons why the option is risky.

Silence.

Laurence had looked up from the job description.

— That's a different competence from what we evaluate in interviews.

— Yes.

— Our evaluation grids don't measure it.

— No.

— And our performance indicators don't reward it either, currently.

It was not an objection. It was a precise observation, formulated with the clarity of a person accustomed to reasoning in systems. Nadia was taking notes. Isabelle was watching Laurence with the attention of someone receiving unexpected help.

What followed took forty minutes. Isabelle tried to formulate. Nadia tried to translate into publishable criteria. Laurence, who rarely spoke at length, intervened in short sequences to try to make Isabelle's request legible enough to appear in a job posting.

It was not simple.

Isabelle's first attempt: someone who supports the decision rather than covers it.

— I can't publish that, said Nadia. It implies that current profiles don't support decisions. No search firm will send us anything.

Second attempt: a lawyer oriented toward trajectory rather than exposure.

— Nobody will understand what that means. And those who think they do will position themselves as strategic counsel. We'll get M&A profiles.

Laurence spoke.

— What if we simply said: capacity to produce an argued recommendation where the law leaves room for manoeuvre?

Nadia re-read the sentence.

— That's the same as a good business lawyer.

— No, said Laurence. — She paused, slightly surprised herself at what she was formulating. — A good business lawyer tells you what could go wrong in a given situation. What Isabelle is looking for is someone who starts from the situation and looks for what the law still permits. That's not the same direction of work.

— And how do we distinguish them in an interview?

Laurence had no immediate answer. She was thinking about something she had not yet had to put into words.

— We submit a situation where the text is dense but not prohibitive. And we see whether their reflex is to map what is possible or to list what is risky.

— They're the same candidates, said Nadia. Depending on how you ask the question, the same profile can give both answers.

— Exactly, said Isabelle.

Nadia set down her pen.

— You're explaining that we're looking for a posture, not a competence.

— Both. But in that order.

The meeting ended without a publishable job description. Nadia had three days of careful work on the table. She left with notes in the margins, two crossed-out attempts and a formulation from Laurence that she didn't yet know how to fit into the standard fields of an ATS.

In the corridor, Laurence stopped for a moment. She was thinking about something that the metrics of her own role did not measure. It was not because this profile was rare. It was because they no longer knew how to read it.

The meeting had started at five to nine because Stéphane always arrived early to sit with his back to the wall, facing the door, and watch people come in with the air of someone who has already won.

He had been running commercial operations for six years. He knew clients by first name, margins by heart, and internal processes the way you know a difficult neighbour: in detail and with resentment.

This morning he had brought a dossier. Not a dossier in the strict sense — a cardboard folder with tabs and a spiral binding — but an eleven-page printout stapled at the top left, which in this company meant: I am serious and I have had enough.

Across from him, Laurence. Head of compliance. Arrived eighteen months earlier with a master's degree in compliance, a certification obtained remotely, and the quiet conviction that her job consisted of preventing the company from doing regrettable things. She had already placed her pen on the table. Cap screwed on. Bad sign.

Between them, a space of four empty chairs that resembled a buffer zone administered by the United Nations.

Marc had just arrived. He had been introduced the previous week at an all-hands as “reinforcement on regulatory matters,” a formulation that had produced three different interpretations depending on which floor you worked on, and none of them enthusiastic.

He sat down in the buffer zone.

Stéphane placed his eleven pages on the table.

— So. The framework agreement with Bremer. Are you familiar?

Laurence nodded. She was familiar.

— We have been working with them for four years. Renewal in March. They want to add a co-development clause on formulations. It is not complicated, it happens in the sector, I have three competitors who signed it without apparent difficulty. And now I am told the legal validation process requires a six-week delay for any new clause on intellectual property.

— That is correct, said Laurence.

— The client wants an answer before the fifteenth.

— I understand the commercial constraint.

I understand the commercial constraint was, in the language of this company, a polite way of saying that is not my problem.

— Laurence. The fifteenth is in nineteen days.

— I can count.

— Then you know we cannot meet the deadline.

— The process exists for serious reasons. A co-development clause creates rights over future creations. It is sensitive.

— Sensitive, repeated Stéphane, as if the word had a strange taste. Bremer accounts for two point two million on the year. If we lose the renewal because we couldn’t respond in time, who explains that to the executive committee?

Silence.

Laurence re-capped her pen. Re-capped it a second time, which was unnecessary but gave her something to do.

— The process was validated by the legal department in 2022.

— In 2022 we had four internal lawyers. Now we have two and a half. The process has not been updated.

— Half a lawyer, said Marc.

Both turned toward him.

— You said two and a half. I am curious which one is part-time.

— It is a figure of speech, said Stéphane, slightly thrown.

— I know. I was asking. — Marc took the eleven pages, leafed through them unhurriedly. — This six-week process. Where does it come from?

— From the legal department, said Laurence.

— I understand where it comes from institutionally. I am asking where the delay comes from. Which text, which rule, which case law requires six weeks to validate a co-development clause in a B2B framework agreement in your sector?

Laurence looked at Marc with the expression of someone who is beginning to dislike the direction a conversation is taking.

— It is not an external legal obligation. It is an internal rule of good governance.

— So it is a choice.

— It is an established procedure.

— Yes. But a procedure established by whom, to protect what, against which specific risk?

The pen was placed back on the table. Definitively this time.

— Intellectual property is a complex area. The delay allows for serious analysis.

— How many clauses of this type have you analysed in the past eighteen months?

Silence.

— Laurence.

— I do not have the exact figure.

— Estimate.

— Two. Perhaps three.

Marc nodded slowly, as if a piece of information were adjusting itself inside something.

— So a six-week delay was established for a risk that presented itself two or three times in eighteen months. And this delay is now potentially costing two point two million in renewal. — He set down the pages. — I am not saying intellectual property is without risk. I am saying the six-week delay comes from no text. It comes from a caution that was transformed into a norm. That is different.

— What you call caution, said Laurence, may have prevented disputes you do not see because they did not happen.

— Possible. Even probable for some of it. But you cannot defend a process on the grounds of what it may have prevented if you are unable to calibrate it against the real risk. The concrete question here is: does a standard co-development clause in a renewal with an existing client of four years present an IP risk level that objectively justifies six weeks of analysis? — He waited. — If the answer is yes, tell me why and I will pass it on to Stéphane that we cannot do otherwise. If the answer is no, then the process has built a wall where the law placed none. And in that case, it is not for Stéphane to explain to the client why we are not responding in time. It is for us to amend the process.

— And if I disagree with your analysis? said Laurence.

— Then we escalate, said Marc. That is what it is there for.

The escalation took the form of an email sent by Stéphane at five thirty in the afternoon, copy Marc, copy Laurence, copy Isabelle Renard, Managing Director, subject: Bremer renewal — arbitration required.

The word arbitration in an email subject, in this company, produced the effect of a faint smell of smoke. No declared fire. Just something that was not right.

Isabelle replied at five forty-four: Tomorrow 8:30. My office. All three.

Isabelle Renard’s office looked out over a car park. She had refused the corner office when she arrived, five years earlier, on the grounds that she did not like rooms where people looked outside while she was talking to them. This detail had circulated through the company as important information that nobody quite knew what to do with.

She was reading Stéphane’s email when they came in. She did not look up immediately. Finished reading, put down her phone, looked at all three of them.

— Summarise the disagreement for me. Not the background. The disagreement.

Stéphane took a breath.

— We have nineteen days to respond to Bremer on a co-development clause. The IP validation process requires six weeks. I think we can move faster without taking serious risk. Laurence thinks the process exists for good reasons and should not be bypassed every time it creates a commercial constraint.

— That is not quite what I said, said Laurence.

— It is what I understood.

— I said the process was established for serious reasons and cannot be emptied of its meaning every time it creates friction.

Isabelle turned toward Marc.

— And you?

— I think the question is not whether to bypass the process. It is whether it is calibrated against a real risk or against a caution that has become autonomous. If the latter, amending it is not emptying it. It is making it work.

— The clause in question. You have read it?

— Yes.

— Risk level.

— Moderate to low on this specific client. The clause is standard in the sector, the rights are well delimited, Bremer has no history of litigation. It is not a clause we are seeing for the first time. — He paused. — The main risk is not in the clause. It is in the fact that if we make a habit of applying disproportionate timelines to low risks, operations learn to bypass the process entirely. Which is considerably more dangerous.

Laurence turned her head slightly. Not quite disagreement. Something closer to recognition of an argument she had not anticipated.

Isabelle took Stéphane’s dossier, opened it to the clause page, read in silence for forty seconds that felt longer to everyone than they probably were.

— Laurence. If I ask you for a serious analysis of this clause in five working days, is that feasible?

— Technically, yes. But it sets a precedent.

— The precedent I want to set is that we are capable of calibrating our timelines against the real level of risk. Not treating a standard clause with the same delay as an acquisition. — She closed the dossier. — Five days. You give me an analysis and a recommendation. Not a list of reservations. A recommendation. Stéphane goes back to Bremer today to confirm we will respond before the fifteenth. — She looked at Marc. — The IP process. I want a revision proposal within a month. Not to remove the delay. To make it intelligent.

Silence.

— Questions?

Stéphane had no questions. He had the look of someone who got what he wanted and is slightly wary of that.

Laurence had the look of someone reorganising something internally.

Marc was looking at the car park through the window.

— One observation, he said.

Isabelle waited.

— This process held for three years without anyone questioning it. Not because nobody was suffering from it. Because nobody knew it was questionable. Operations thought it was the law. Laurence thought it was necessary caution. Management had not heard about it because Stéphane had until now been absorbing the friction himself. — He paused. — The wall had been invisible for three years. That happens often. It is not negligence. It is how processes die: not in a grand disagreement but in a silent accumulation of small workarounds and lost clients whose departure nobody quite understands.

Isabelle looked at him for a moment.

— That is why you are here, she said.

It was not a question. Marc did not answer. Which was, in the circumstances, the correct answer.

Laurence’s analysis arrived on Thursday at four fifty-eight, two minutes before the tacit expiry of the five working day deadline Isabelle had set.

Marc had worked with her for the last two days. Not at her request. He had appeared on Tuesday morning with the clause text printed out, three court of appeal decisions on co-development disputes in the chemical sector, and a question: What exactly are you trying to protect with the six-week delay?

Laurence had initially had the look of someone being asked to justify the existence of gravity. Then she had answered. What she was trying to protect, in reality, came down to three points: ensuring that existing company rights over formulations were properly bounded before any new clause; verifying the clause did not create an implicit disclosure obligation on ongoing developments; and ensuring the exit mechanism was workable if the commercial relationship deteriorated. Three points. Estimated at six weeks.

Marc had looked at the three points.

— The first can be resolved with a definition in the clause itself. Two sentences. The second — there is consistent case law since 2019 that delimits this point very precisely in your sector. It is not a grey area. The third — Bremer accepted a termination for convenience clause in the current framework agreement. It can be carried over with one adjustment.

Laurence had looked at her own notes.

— The six-week delay was set in 2022 by my predecessor.

— I know.

— There had been no discussion about what cases it applied to.

— I know that too.

— So we applied six weeks to everything because six weeks had been set once for something complex.

— Yes. — He paused. — It is fairly common. It is not a criticism.

— It sounds like one.

— It is an observation. Processes fix themselves to the most difficult case encountered and stay there. Nobody revises them because revising takes time and the status quo costs nothing visible — as long as Stéphane absorbs.

— And when Stéphane no longer absorbs.

— When Stéphane no longer absorbs, we have a meeting at eight thirty with the managing director.

Laurence had smiled for the first time that week. A brief, almost involuntary smile, the kind that escapes before you have decided to let it pass.

— The analysis, she said. How should I write it?

— As a recommendation. Not a list of what could go wrong. Isabelle asked for a recommendation and she meant it. — He paused. — State what you found, state what you recommend, state why. If you have real reservations, one or two sentences. Not eight pages of caveats.

— And if it goes wrong?

— If it goes wrong, the analysis you will have written shows you examined the risk points seriously, identified reasonable protections, and formulated an argued recommendation. That is defensible. — He took his coat from the chair back. — What is not defensible is a six-week delay applied mechanically without being able to explain what it was responding to.

He had left. Laurence had looked at her screen. Then she had written a recommendation. Four pages. Clear conclusion. Reservations: two sentences. She had read it three times, looking for what was wrong. Nothing was wrong. It was simply shorter than what she was used to producing, and this brevity gave her the feeling of a lightness she was perhaps confusing with imprudence. She sent it at four fifty-eight.

Bremer signed the renewal on the nineteenth of the month. Stéphane sent an email to the team with the subject Done and no further text, which was his way of marking victories without appearing to care about them.

Marc spent the following week mapping the other contractual validation processes. He worked the same way in each case: find the originating text or rule, understand the initial risk that had motivated the delay or procedure, measure the gap between that risk and the current application. What he found, in most cases, was less scandalous than ordinary. Reasonable processes at the time of their creation, never updated, applied uniformly to increasingly heterogeneous situations by people who had neither the mandate nor the time to question them. Nobody had decided to slow the company down. The accumulation had done it in their place.

He noted everything in a document he called, for lack of a better term, state of the terrain. Not a cartography in the graphic sense. A list of questions with provisional answers: where does this wall come from, what does it cover, what does it cost operationally, how can it be amended, and by whom. Some walls were real. A sectoral regulatory clause that genuinely imposed a notification delay. A legal review obligation for contracts above a certain threshold, written into the articles of association. Those walls Marc documented as such, without trying to work around them.

That was the distinction he tried to explain to Laurence one Wednesday morning at the building’s café.

— What I am looking for is not to remove constraints. It is to distinguish those that exist from those we have manufactured. If it is the second type, amending it is not emptying it. It is making it function.

— And when we have manufactured them, said Laurence, does that mean they were useless?

— Not necessarily. It means they can be questioned. A legal constraint, you adapt to. An internal constraint, you evaluate. The confusion between the two costs enormously because it puts on the same level what is negotiable and what is not.

Laurence looked at her coffee.

— My predecessor thought his role was to protect the company against itself.

— That is a legitimate role.

— You do not really think that.

Marc took a moment.

— I think protecting the company against itself is useful when you are protecting the decision from its own excesses. It is not the same thing as protecting the company against its own trajectory. In the second case, you are no longer protecting. You are governing in place of someone who was chosen to govern.

Laurence set down her cup.

— That is why you asked me to write a recommendation and not a list of risks.

— Yes. A list of risks says: here is why you might be wrong. A recommendation says: here is what I think, and I assume the coherence of it.

— That is more exposed.

— It is more honest. And it is defensible differently. If Bremer attacks us in two years over the clause, the dossier shows we examined the tension points seriously and retained the most sustainable option in view of the real risk. That is not cover. It is a record.

— A record.

— Of the reasoning. Not of precautions. — He finished his coffee. — The difference, before a judge, is considerable.

The regulator announced itself on a Tuesday in November, in the form of a recorded delivery letter addressed to the managing director, subject: request for information in the context of a sectoral review of co-development contractual practices. Not a prosecution. Not a summons. A seventeen-point questionnaire accompanied by a request for documentation on contracts signed in the category over the past three financial years.

Isabelle convened Marc, Laurence and the company’s usual external counsel, a man named Gilles who arrived with an associate, a Moleskine, and the contained energy of someone who bills by the hour.

— Standard situation, said Gilles after reading the letter. Sectoral review, they are casting wide. This probably does not target the company specifically. We need to respond seriously, within the deadlines, and produce nothing beyond what is requested.

— We have the contracts, said Isabelle.

— We will need to contextualise them. Explain the contractual policy, the validation procedures, the decisions made.

— We have that, said Marc.

Gilles turned toward him with the slight surprise of a man who had not yet placed all the people in the room.

— What exactly do you have?

— For co-development contracts over the last eight months: the clauses, the analyses produced before signature, the recommendations retained, the options set aside and the documented reasons. For the three financial years, it is less systematic for the earlier period. We will need to reconstruct some of it.

Gilles looked at Isabelle.

— You have documentation of your contractual arbitrations?

— Since approximately eight months, said Isabelle. We changed method.

— That is unusual.

— Probably.

Gilles made a note in his Moleskine. The associate made the same note in his. For the earlier period, the reconstruction would turn up reservations, delays, conditional validations — not much that explained why they had decided what they had decided.

— That is the case in most companies, said Gilles, with the neutral tone of a man who has spent twenty years not judging his clients.

Marc was looking at the regulator’s letter on the table. He read in silence for two minutes.

— Half the questions have a direct answer in the documentation from the last eight months. The other half concerns the earlier period. On that period, we can reconstruct the facts. What we will not easily reconstruct is the intention.

— The intention, said Gilles.

— Why we retained one clause rather than another. What level of exposure had been considered. Which alternative was set aside and for what reason. Without that, the contracts exist but are silent on their own logic.

— Contracts do not have to explain their logic, said Gilles.

— No. But when a regulator asks to understand a contractual policy, contracts without any record of their arbitrations look like a policy without an author. It is not illegal. It is simply harder to defend coherently.

Gilles set down his pen.

— You are right. It is simply rare to hear it put that clearly before the situation becomes serious.

— That is the advantage, said Marc. We are not yet in the serious.

Isabelle looked at the table.

— For the earlier period. What do we do?

— We respond with what we have. Factually, without overinterpretation. We do not fabricate a retrospective logic we cannot sustain. — Marc set down the letter. — And we document the next eighteen months very carefully.

— That does not protect us today.

— No. It prepares us for the next time. — He looked at Isabelle. — There will be one.

It was not a threat. Not a prophecy. Just information delivered with the tranquillity of someone who has long accepted that risk does not disappear. It is governed.

Gilles closed his Moleskine. The associate closed his.