News in Practice
“You are faster now, so it should cost less”
The demand is now raised in almost every relationship between a legal department and its counsel. It is partly founded, and it redefines a good deal more than the size of an invoice.
For lawyers whose clients are opening this conversation, and for the legal departments opening it. What the hour measured, what it has stopped measuring, and what is at stake between the two beyond price.
The question no longer comes from firms wondering what to do about artificial intelligence. It comes from their clients, who have done the arithmetic for them.
Several large American banks, Goldman Sachs, Morgan Stanley and Citigroup, now ask their counsel to demonstrate how efficiency gains from artificial intelligence show up on their invoices. Citigroup requires firms bidding for a mandate to quantify their savings. Morgan Stanley is moving a large share of its external work to competitive tenders and alternative fee arrangements, and the group's general counsel calls the way firms are paid “extraordinarily unstable”. The reasoning fits in one sentence: if the work takes less time, the bill should come down.
That sentence is true and incomplete. What it misses is not a subtlety of billing, it is what the hour had been measuring all along.
What the demand gets right
Start with what the demand gets right, because it is right about the essential point.
Part of legal work has always been paid at its cost of production. Reviewing a stock of contracts, or producing the first version of a standard instrument: the client paid for hours because those hours had to be spent. If they are no longer needed, the client is entitled to notice it and to say so.
The demand is not new either. Fee caps, fixed fees and alternative arrangements have existed for twenty years. What has changed is the argument: the client is no longer discussing its own budget, it is discussing its supplier's cost structure. That is a far harder position to refuse.
So the point has to be conceded, not negotiated. For the work whose cost the hour measured, the hour no longer measures it.
What the hour was really measuring
The rest of the work was never paid at its cost of production, and that is where the sentence cracks.
When a client pays for a partner who has negotiated two hundred shareholders' agreements to read one, the client is not paying for forty minutes of reading. It is paying for the probability that the clause which will cause trouble in three years is caught this morning. The hour served as a unit of account for something that was not time: a judgement, and the responsibility that comes with it.
As long as the two were mixed in the same invoice, nobody needed to tell them apart. Artificial intelligence separates them brutally, because it compresses one without touching the other.
Hence the real difficulty of the conversation. Both parties are talking about the same matter, but one is pointing at the cost of production and the other at the value of the commitment. Neither is lying.
The banks leading this movement know it, as it happens. What they are opening to competition and shifting to alternative arrangements is the work whose volume can be measured. For the counsel you call when a deal starts to wobble, they will go on paying, and paying well. The demand for a cut only bears on half the invoice, and the firm still has to be able to show where the line runs.
Three possible responses
This discussion does not call for a single pricing response.
Three responses can be distinguished: change nothing, pass the gain on, or bill the outcome rather than the time. This is our analytical framework, not a measure of the market's distribution.
Each of these positions rests on a different reading of what the client is buying, and each can be defended.
Change nothing
A first position is to absorb the gain without changing the price.
It holds up where the fee was already set on the value of the matter rather than on a count of hours, which is the case for much of strategic advice and high-stakes litigation. The client was not buying hours, so it has nothing to recover.
It becomes untenable the moment the client sees the line. A detailed time breakdown sent to a legal department that knows what reviewing a stock of contracts costs today produces the opposite of the intended effect. That is precisely what the banks cited above have begun to do: ask for the detail.
Lower the price of the work
A second position is to pass the gain on.
It is coherent and it has a cost. Passing a fall in production cost through to an hourly fee amounts to confirming that what was being sold was time. The firm wins this year's conversation and loses the ground on which it will hold the next ones: once it is established that the price follows the hours spent, every technical advance reopens the negotiation, and always in the same direction.
It makes sense in one case: where the work really is volume, and the firm chooses to treat it as such, separating it from the rest.
Bill the outcome rather than the time
The third position is the most instructive, because it contradicts the general intuition.
Charging more for work assisted by artificial intelligence only makes sense if the unit of account changes. A higher hourly rate applied to fewer billed hours describes exactly that: the firm is no longer selling duration, it is selling what an hour now produces.
The clearest form of this position is the fixed fee with a deadline attached. The client knows the price and the date before the work starts, the firm carries the execution risk, and the number of hours leaves the conversation because it is no longer part of it. What is sold becomes an outcome delivered by a date, which is precisely what the client was buying without being able to put it into words.
This position assumes two things, and collapses without them. That the firm knows what the matter actually costs it, failing which the fixed fee is a lottery. And that it owns, in front of the client, the use it makes of artificial intelligence, since a price that no longer follows the hours cannot be explained any other way.
What no billing model settles
Whichever model is chosen, one question remains open, and it determines the value of everything else.
Who answers for what is delivered. A lawyer will never be able to say that the error came from the tool. The instrument carries their signature, their professional liability is engaged, and no fee arrangement displaces that commitment. A legal department that obtains a discount without knowing who checked what has not struck a good bargain, it has transferred a risk it cannot measure.
That is why a negotiation conducted on price alone is a poor negotiation. The useful question is not how many hours were spent, nor even what the matter cost. It is what was checked against which sources, and what will remain of it as a record if someone asks in eighteen months.
A firm able to answer that question holds its price. A firm that cannot will defend it less and less well, whatever its billing model.
What remains when the wave subsides
What remains is what this conversation leaves aside, and what will nonetheless decide the relationship.
A client does not choose its counsel on a rate card. It chooses them because it knows how they react when a matter becomes complicated, and what it will get when it calls on a Friday evening. It also chooses them because it does not have to re-explain its organisation with every request. That mutual knowledge cannot be bought and cannot be automated.
The present period pushes it into the background, because everyone is looking at the tool. It will come back, for a simple reason: when a capability becomes available to everyone, it stops distinguishing anyone. What will then set one adviser apart from another is what set them apart before, and which the wave has merely covered over.
The client's demand is therefore legitimate and badly framed. The answer it calls for is not a percentage, it is what the money now buys.