Industry
The Day Saving Time Starts Costing Money
AI does not always turn saved time into revenue. Everything depends on what the firm has chosen to sell.
For two centuries, one equation governed the modern economy almost without fail: raising productivity produced more wealth. Doing things faster, more efficiently, with less, translated sooner or later into more value created and captured. This equation was so constant that it came to seem natural, almost a law. Yet artificial intelligence may be the first mass technology to make this contradiction so visible in the trades that sell their time, where saving time can, in certain economic structures, produce not more wealth but less. The rupture is not new in itself, other productivity gains had already, elsewhere, destroyed revenue; but never at this scale nor at the heart of activities that bill time itself. This is not a sectoral detail. It is a crack in the link once thought unbreakable between efficiency and value.
To understand why this equation breaks, one must look at what it silently assumed. It held as long as the efficiency gain could be sold: producing twice as fast enriched whoever could sell twice as much, or lower prices to take the market. The link between efficiency and wealth ran through the ability to turn time saved into volume sold or competitive advantage. This mediation still works in industry and most services. But it seizes up wherever what is sold is indexed, not on a volume, but on time itself. There, saving time does not free up capacity to sell: it destroys the unit of what was being sold.
Efficiency enriches only if it can be sold. AI is the first technology to hit that condition head-on.
Where the equation breaks first
Law firms are one of the places where this rupture shows most clearly, because they embody, almost in pure form, the economy of sold time. A firm that bills its hours literally sells time; its wealth is a volume of hours multiplied by a rate. In this structure, work that took ten hours and now takes three produces an obvious benefit for the client and relief for the lawyer, but it mechanically cuts the firm’s billable base. The technology that promises efficiency threatens, in the same gesture, the very unit part of the revenue rests on.
The obstacles to AI adoption are usually presented as technical, organizational or cultural. They are. But there is another, deeper, that is neither fear nor habit, but well-understood self-interest: why would an organization fully adopt a technology that erodes the unit on which it bills? This obstacle is not a resistance to correct with pedagogy. It is the local translation, in a firm, of the general economic crack: efficiency ceases to enrich there because what is sold is the time efficiency precisely makes disappear.
What makes this case exemplary is that it is not isolated. Every trade that bills time, from consulting to audit, will know the same tension as AI enters it. The law firm is only where the phenomenon is most legible, because the profession made billed time an almost universal norm. What plays out there is a prefiguration: wherever price has been moored to time, AI will force that mooring to be undone, or the erosion to be suffered. Law is ahead in this realization, not by virtue, but because the contradiction is more naked there than elsewhere.
Wherever price has been moored to time, AI will force that mooring undone, or the erosion suffered.
What the current debate does not measure
The debate on AI abundantly measures certain things: hours saved, speed of execution, volume produced, cost of the tool. These figures are real and easy to display. But none answers the one question that decides the value created: who captures the time saved? A time saving is not, in itself, an economic gain. It becomes one only if someone, in the chain, turns that freed time into retained value. As long as one does not ask this, one measures an efficiency without knowing whether it enriches or impoverishes, and one falls back precisely into the illusion that the productivity-wealth equation would always hold on its own.
Yet the answer depends on a variable the technical debate ignores: the billing model. The same time saving produces opposite effects depending on how the work is sold, and it is this variable, not the tool’s performance, that decides whether AI is an asset or a threat. Two firms with the same tool, equally efficient, can see one’s revenue grow and the other’s melt, for the sole reason that they do not bill the same way. The technology is identical; the economic model decides its sign.
A time saving is not an economic gain. It becomes one only if the economic model knows how to capture it.
Three models, three fates
Take the three ways of billing and follow what efficiency does to each. At the hourly rate, the logic is cruel: a matter’s revenue is time times rate, so reducing time reduces revenue, unless the rate rises as much, which the market does not always allow. Under this regime, AI’s efficiency works against the firm: the more efficient it is, the more it cuts the billable base. This is where the contradiction is most naked, and it is also the model still dominant across much of the profession.
At the fixed fee, the logic inverts. The price is set in advance, independent of the time actually spent, so every hour saved turns directly into margin. AI then becomes a powerful profitability lever, but on one strict condition: that the firm priced risk and complexity correctly when setting the fee. Poorly calibrated, the fixed fee transfers to the firm the risk the hourly model placed on the client, and AI’s efficiency does not recover imprudent pricing. The fixed fee rewards efficiency, but only if the pricing is right.
At value-based pricing, finally, the contradiction disappears. The price is no longer indexed to time consumed, but to the value produced for the client, the scope of the stakes, the criticality of the decision. The time AI saves then diminishes nothing, because time was not what was being sold. Efficiency becomes a pure benefit, captured without erosion of the base. It is the only one of the three models where saving time and making money cease to be in tension, and it is no accident that it is the one toward which the most lucid firms are beginning to move. There, restored, is the old equation: efficiency enriches again, because the value sold is no longer time.
It is not the tool that decides whether efficiency enriches or impoverishes. It is what one has chosen to sell.
Selling something other than time
One then sees that the transformation AI calls for is not technological, it is economic, and it reaches far beyond law. What AI imposes, on every activity that billed time, is a clarification: what am I selling, at bottom, when I can no longer sell the hours the machine made disappear? Law faces this question earlier and more nakedly than others, but it does not face it alone. The answer, everywhere, takes the same form: to stop selling the input, time, and to sell the output, the result, the security, the judgment, the mastered risk.
This shift is difficult, and it would be dishonest to present it as a simple adjustment of the rate card. Selling value rather than time supposes knowing how to name it, defend it, make it accepted by clients used to reasoning in hours. It is commercial and cultural work as much as economic, and it is not decreed. But the direction is clear, and AI only makes urgent a movement the most advanced organizations had already begun. Technological efficiency forces a whole profession to say what it really sells, and this long-deferred clarification suddenly becomes vital.
For the paradox turns into an opportunity for whoever understands it early. Whoever stays on time will suffer AI, caught between clients who will demand the price cut efficiency makes possible and a billable base that shrinks. Whoever has shifted their model toward value will reap efficiency as a net gain, and can invest it in what billed time never financed: quality, prevention, the relationship, security. The same technology impoverishes some and enriches others, and the variable is not the technology, it is the model one will have had the courage to change before being forced to. The productivity-wealth equation is not dead; it has become conditional, and the condition is the model one chooses.
The firm that wins with AI will not be the one that works fastest. It will be the one that learned to sell something other than the time it saved.
It is this clarification that MAX seeks to make possible rather than to impose. A layer that makes efficiency measurable, traceable, governable gives the firm the means to know what it really produces, and therefore to price it other than by the hour. Technology does not resolve the economic question alone, none does, but it gives the instruments of a model where the value created can finally be seen, defended and sold for what it is. The rest, the decision to sell something other than time, belongs to the firm, and it is the most important decision the profession will have to make in the decade.
