In most teams there is one person who has read all three documents. Not the summary, not the deck — the meeting notes, the old contract annex and the ticket somebody opened in March. That person is rarely the most senior in the room, and there is no qualification anywhere that says what they do.
There are certificates. There is a standard with a number. There are books with the word “engineering” on the cover. There is no course at the end of which somebody says: right, you can do this now.
The job that was never a job
In the nineties it had a name and a chair: business analyst. The chair sat between the business and IT, and the work went both ways — turning “we need this by the end of the quarter” into a document, and turning the document into a question somebody could actually answer. The document was thick, it was signed, and it was out of date by the third change request.
Then the translator's chair was recognised as what it also was: a place where information stopped moving. So it was removed. What remained of it landed on the Product Owner — a role that combines vision, prioritisation and market judgement on paper, and in practice is often whoever writes the tickets. Not because they trained for it. Because there is nobody else.
The names come back periodically, the way fashions do: requirements engineer, business analyst, product manager, delivery lead. Behind them, the desk has barely changed in thirty years. An inbox of wishes that are not requests. A meeting where three people use one word for three different things. A list, and the private knowledge that half of it is not sharp enough to build from. And the near-certainty that nobody will notice, as long as you keep noticing.
Done well, it looks exactly like a week in which nothing happened.
Why it read as administration
Three reasons, and none of them is contempt.
Start with the one above, because it is the most durable. This work is invisible when it succeeds. A sentence that was exact enough costs nobody a day. Two days that never happened appear in no report. What appears is the other case — the escalation, the module built against the wrong assumption, the date that slips. A discipline whose success consists of absent events gets no stage. It gets asked why it is taking so long.
Then the bottleneck. While weeks stood between an idea and something running, building was the expensive step, and anything that is not the bottleneck reads as overhead in every organisation ever drawn as a chart. That is not a judgement about the people. It is arithmetic, and it was correct for as long as its premise held.
And third: no workpiece. You recognise a craft by the thing that stands there at the end — something you can look at, test, and hand to somebody else. This work never had one. Its output was a document that got read once and was then overtaken by the code. Where two truths exist and one of them executes, the executable one wins every time. So the document became paper within a month, and the work that produced it looked, in hindsight, like effort without an object.
What moved
The bottleneck moved. Give an agent a clean brief today and a few minutes later something starts, with tests, in a shape you can take apart. Not everywhere, not inside every system that has been growing for fifteen years, and not without somebody who reads what came back. But across enough of the work that the arithmetic no longer holds. What decides the result is the brief — and a brief is a sentence somebody wrote.
Which puts this activity, for the first time, at the point where things are decided rather than upstream of it. Why that is a shift in authorship rather than a new tool chain is the subject of The sentence is the artifact. What matters here is the other half of the question: what does good look like, in a discipline nobody was ever taught?
What good looks like
Not diligence. Four things you notice in somebody who can do this.
They see the soft sentence before anybody else reads it. “More flexible”, “more performant”, “more user-friendly” — these are not requirements. They are honest summaries of something nobody has finished thinking through. Spotting them is practice, not talent, and it is the same reflex a copy editor has on the third read of a paragraph that sounds fine.
They know which question is the expensive one. Every unclear sentence has ten possible follow-up questions. Nine cost five minutes and change nothing. One changes the scope. Finding it without asking all ten is the difference between a twenty-minute call and a two-hour workshop that nobody accepts the invitation to next time.
Their sentence works without them. The test is cheap: hand it to somebody who was not in the meeting and watch what gets built. If you have to stand next to it and fill in the gaps, it is not finished — half the meaning is still in your head, and your head goes on holiday in August.
They know when to stop. Twenty sharp sentences beat two hundred soft ones, and the two hundred are the more comfortable work, because volume feels like thoroughness. Stopping without the sense of having left something out is the last of these four to arrive, and the one people mistake for carelessness.
There is no syllabus
If there were one, it would not open with principles. Principles can be learned in an afternoon and applied in no year at all. It would be a catalogue of failure modes, the way every real trade teaches: the sentence that describes a feeling instead of a result. The sentence you cannot read an actor out of. The requirement created twice because one customer phrased the same wish two ways, six months apart. The scope that lists what belongs and never what expressly does not. What it looks like once that chain has snapped is Why requirements fail.
The reason no such syllabus exists is unglamorous: this was never a profession, it was a leftover responsibility. So it is learned the way trades have always been learned — from somebody already doing it, on a real piece of work, with a correction that stings. Only without a workshop, without a certificate, and usually without permission to take the time.
That is the actual point of this piece. If you have read this far while thinking of one specific Thursday — a sheet with forty rows of which six are right, and the feeling that somebody properly trained ought to be doing this instead of you — then it is worth saying plainly: that person does not exist. There are only people doing the same work, who have made the same mistakes twice.
What has changed since ten years ago is not that the work got harder. It is exactly as hard as it was, and it has stopped being inconsequential. What you write down gets built — faster, closer to the wording, and with fewer questions coming back than at any point before. That is not a promotion. It is the end of the excuse that the phrasing does not much matter.