Once, a well-presented document signalled something. Not that its content was right, but that someone had spent time on it, and that one could therefore reasonably assume they had also thought about the substance. Form was a clue to care. That clue has become free, and we have not yet found what to replace it with.
We have addressed the question of merit and that of a text's authenticity. Here is the question that precedes them all: how do we recognise work done well?
What we were really judging
We almost never have the means to verify work in detail. A client does not reread every line of a report, a reader does not check every claim, an employer does not examine every decision.
So we proceed by clues. Impeccable spelling, a clear structure, careful formatting, a complete bibliography. None of these elements guarantees the correctness of the substance. They guaranteed something else: that someone had invested time.
And that investment was correlated with seriousness. Not perfectly, but enough to be relied upon. Work that was sloppy in form was generally sloppy in substance too.
What AI removes is not quality. It is the correlation between visible signs and effort invested.
The shift under way
The consequence is that evaluation criteria are shifting towards what remains costly to produce.
Verifiable accuracy. An exact figure, a source you can open, a claim that holds up under scrutiny. We saw this with lawyers sanctioned for invented case law: the form was perfect, and that is precisely what misled.
Contextual relevance. Well-done generic work is now commonplace. Work that takes account of the specific situation, the real constraints, what happened last week, presupposes knowledge that no tool has.
The ability to defend. Being able to explain why this choice rather than another, what was set aside and why. This is the criterion we proposed for authenticity, and it works here too.
There is a tradition of thought about work done well that has never made quality depend on difficulty. An experienced craftsperson works faster than a beginner, and better. Mastery reduces visible effort while increasing quality. From this point of view, a tool that removes toil degrades nothing: it simply shifts where skill resides. What distinguishes remains judgement, not sweat.
The loss that is real
But we must name what is lost, otherwise the analysis would be complacent.
Work done well was not only a result; it was also a relationship to what one does. Searching for the right word, reworking a formulation three times, redoing a demonstration because it does not quite hold: these gestures produce a satisfaction that the result alone does not give.
We wrote this in our article on time saved: certain efforts transform the person who performs them. Delegating these gestures does not necessarily degrade the product, but it changes what the work does to the person doing it.
This loss is not measurable and does not appear in any productivity report. Yet it is what makes a trade worth practising.
What to take away
The question has no universal answer, and it arises differently depending on whether one is judging the work of others or one's own.
To judge another's work, clues of form no longer serve. You must look at what still costs: verifiable accuracy, adaptation to the specific case, the ability to answer for one's choices.
To judge one's own, the criterion is more intimate and more demanding. It may come down to this question: is there still a part of what you do that you would not have wanted to delegate, even if you could have? If the answer is no, the work may be well done, and it has probably ceased to be yours.