A friend of mine is a senior translator. He has spent decades building expertise in turning complex legal documents from one language into another. Not just the words, but the intent, the register, the cultural weight behind them. Another aspect of his work is consistency: making sure that the same concept is rendered the same way across every document in a case or a product family, a kind of invisible threading that holds legal and technical language together. When we met recently, I asked him how AI was changing his work.

His answer was that he is spending more and more of his time supervising AI, and less and less actually translating, or mentoring the juniors who used to sit beside him. His employer is not just accepting AI-generated translations. They are actively encouraging the team to rely on them. And the reason is not laziness or ignorance. It is economics, and the economics are extraordinary.

Here is the conclusion I have reached from that conversation: AI has not disrupted the translation market in the way we usually mean by disruption. It has created an entirely new point on the map, so far from everything that came before that the geometry has changed. I think the right response is not to resist this. But it does require us to make a deliberate choice about what we protect, and why, before the economics make that choice for us.

A point that never existed before

Think of the translation market as a spectrum with price on one axis and quality on the other. For decades, the variables were cost and proximity to expertise, geography, seniority, specialisation, years in a particular niche. Companies were always choosing how far from the ideal they could afford to stray. A senior translator in Geneva cost perhaps two or three times more than a competent one in Bogotá. The quality difference existed, and so did the supervision overhead. Companies weighed those trade-offs consciously.

AI has not moved along that spectrum. It has created an entirely new point on it, one that simply did not exist before. The cost differential is no longer a factor of two or three. It is a factor of a thousand or more. And the quality, while imperfect, is reasonable. Good enough for most purposes. Good enough that a rational management team, looking at their costs, would have to work hard to justify not using it.

This is the thing that vague statements about AI "disrupting" industries miss. It is not disruption in the sense of a better competitor entering the market. It is the appearance of a coordinate on the map so far from all the others that it changes the geometry entirely.

Why no one is listening to the critics

When the economics shift this dramatically, something predictable happens: everyone converges on the new optimum. Economists call this a Nash equilibrium: a point where, given what everyone else is doing, no individual actor has an incentive to deviate. Once one firm normalises AI-assisted work, its competitors face a choice between following or accepting a permanent cost disadvantage.

What we are actually giving up

My friend mentioned something almost in passing: in his industry, the emerging convention is to designate one version of a document, usually the English one, as the canonical reference in case of any clash between language versions. This is a simplification. A deliberate, practical, slightly uncomfortable simplification. Instead of pursuing perfect legal equivalence across every language, you pursue reasonable quality everywhere and designate a tiebreaker.

This is what AI-driven change looks like in practice. We are not losing quality randomly. We are making an explicit choice: the last twenty percent of quality, the part where mastery lives, is no longer worth competing for.

And the old competitive race, companies striving to attract and retain people willing to go that extra distance, is what funded and sustained the whole system. The last twenty percent was not just a differentiator. It was the engine that built everything beneath it.

AI has not simply offered a cheaper option. It has created a new race, on a completely different part of the spectrum. The premium now is on cost efficiency, not excellence. The race has moved. The incentive to compete at the quality end has collapsed. Quality is still possible, but the new point on the spectrum has made the economics of that race unattractive.

Embrace the simplification. Protect the mastery.

The simplification is real, the economics are decisive, and pretending otherwise helps no one.

But the twenty percent that remains, the part where nuance lives, where judgement is irreducible, where a human being with decades of experience notices something a model cannot, that deserves deliberate protection. That level of mastery may not always be commercially necessary. But someone has to know what excellent looks like, or we lose the ability to recognise when the eighty percent falls short.

The race to the quality end of the spectrum built the world we inherited: the professionals, the institutions, the standards we now take for granted. The new race will build a different world. What we owe to the future is at least an honest recognition of what we are choosing to stop competing for, and why it mattered.