Nobody Adopts Efficiency. Everybody Adopts a Raise.

Take "efficiency" as it is pitched to you and run it through your own book. On your fixed-fee matters, the savings become margin the client never sees: invisible, unattributable, and undefended on the day anyone looks. On your hourly matters, the hour converts the savings into a revenue cut on your own books · the mechanism this series derived in "The Partner Without a Computer Was Reading His Balance Sheet Correctly," cited here as settled ground · so passing them through is fee competition you start against yourself. That is the entire menu as the vendors present it. A margin you must hide, or a discount you must fund. Two trades, and no rational partner signs either, which is why thirty years of exhortation produced thirty years of refusal.

The decomposition has a third branch, and it is the only one anybody has ever actually adopted.

1 · Two trades nobody wants

Spell the decomposition out, because as far as we can find no published piece states it, and everything else in this memo hangs on it. "AI efficiency," landed at the individual lawyer's desk, is not one proposition. It splits on the paper the matter is billed under.

Under a fixed fee, efficiency is margin. The work costs you less to produce; the client pays the same; the difference accrues silently. It feels like the good trade until you ask what defends it. Nothing defends it. It is not priced, not disclosed, not attributable to anything the client can see, and the first procurement review that asks how the fee was built finds a number with no floor under it. You are not holding a margin. You are holding a story about a margin, and stories about margins have a way of being repriced by whoever audits them.

Under the hour, efficiency is worse than undefended: it is self-inflicted. Every unit of work the technology compresses out of billed time arrives on your books as a smaller invoice with your own name on it. The earlier memo in this series prices that conversion fully and we do not re-run it; one sentence suffices here. The hour taxes the payoff itself, so the rational partner declines the payoff.

Two branches, both losing. The slow-walk was never a character flaw; it was the correct reading of this menu. Which is what makes the third branch interesting: it is not a compromise between the two trades. It is a different transaction.

Meanwhile, the trade press has already announced the ending.

2 · The headline is printed, and the buyer is already calling its bluff

"AI raises the price of the hour rather than killing it" is not this memo's claim. It is consensus, in print at least four ways. Jeroen Villé, co-founder of Alice, published the scenario in October 2025 under the label "Premium Billable": "higher hourly rates, fewer hours," with the hour evolving into "a signal of expertise, strategic value, and trust." Michael Grupp, CEO of BRYTER, made the same call in Bloomberg Law that December: "The result is a more productive lawyer whose hours are worth more, not fewer." Sean Fitzpatrick, CEO of LexisNexis North America and UK, went further on a Legalweek panel (via Business Insider and TechSpot): "It doesn't take that much inflation to get to the $10,000-per-hour billable hour." And LawFuel's May 2026 examination of the billable hour observes that firms are already "absorbing the productivity gain into higher effective rates, expanding scope, and pushing the saved hours onto more matters." Even the genre's standing complaint · the $2,000-hour problem, the running argument over whether any hour can be worth that · concedes the direction of travel: the reporting around Fitzpatrick's line notes top senior partners already billing close to $2,100.

So the open question was never whether the rate rises. It is which version of the rise survives contact with the client, because the client has started auditing.

Villé's version, read closely, rests on narrative. The hour becomes "a signal of expertise, strategic value, and trust": the rate rises because the lawyer says the hour is worth more, and the client is asked to believe it. Here is what the buyer side says to that. The Association of Corporate Counsel and Everlaw surveyed 657 in-house legal professionals across 30 countries in October 2025, and the headline finding is a bluff being called in real time: "59% report no GenAI savings from their law firms yet." Only 24 percent are satisfied with how outside counsel use generative AI for cost reduction. Sixty-one percent intend to push for changes in how firms deliver and price services. The renarrated rate rise · the one defended by adjectives · is being inventoried by the people who pay it, and they are not finding the value the story promised.

One adjacent market has already run this film to the final scene.

3 · KPMG, or densification without an ops layer

In February 2026, the Financial Times reported, and TechSpot carried, a negotiation worth reading twice, because the roles are not what the headline suggests. KPMG was not the provider in this story. It was the client. "KPMG International urged Grant Thornton to share any cost savings from its AI rollout across its audit operations": KPMG, buying audit services from Grant Thornton UK, demanded its supplier's AI savings be passed through, and when Grant Thornton resisted, KPMG reportedly warned it might take its audit elsewhere. The fee fell, in the figures reported, from $416,000 to $357,000 · roughly 14 percent, surrendered.

Sit with the role reversal, because it strengthens the lesson rather than softening it. One of the world's four great professional-services houses, an organization that knows precisely what AI does to a cost base because it is doing the same thing to its own, walked into a fee negotiation with the discount pre-computed. Grant Thornton had presumably absorbed its AI gains the way Villé recommends and LawFuel observes: into effective rates, quietly. It did not matter what story accompanied the fee, because the client was no longer buying the story. The provider claimed AI productivity but could not show delivered, measured density · and so the client took the rate move for them, in the client's direction.

That is densification without an operations layer: the gain is real, the measurement is absent, and whoever measures first sets the price. The sophisticated buyer of professional services now arrives with the discount already computed. Legal's version of that meeting has a published calendar.

4 · LawFuel's calendar, answered by name

LawFuel's piece is the strongest prior art this memo faces, and it deserves a direct answer rather than a nod. Its scenario, quoted exactly: the hourly model's reckoning comes "probably not in 2026. Probably not in 2027, either." It lands "somewhere in the 2027 or 2028 procurement cycle," because "every GC who deploys AI internally builds a benchmark for how long a task should take. That benchmark becomes the negotiation anchor for outside counsel work." The trigger LawFuel nominates is concrete: a public fee dispute at the Am Law 50 level that turns AI invoice auditing from a procurement curiosity into "a board-level expectation." (One housekeeping note: "densification" is our word for what firms are doing to the hour, not LawFuel's; the framing of that piece is its own.)

Concede the calendar in full. The audit is coming; the GC benchmark is the buyer's version of the KPMG move, generalized and institutionalized; nothing in this memo argues the clock stops. LawFuel is right about the timer. It is wrong about the victim.

The GC's benchmark anchors hours per task. What it kills is the unmeasured rate rise: the hour whose new price is defended at renewal with narrative, the Grant Thornton position. What it rewards · what it certifies, matter by matter · is the hour that beats the benchmark on total cost while carrying a higher rate, because the benchmark measures exactly what that hour is built to show: how long the task took, what it cost the client end to end, what was delivered. A benchmark is only a threat to a provider with nothing underneath the number. To a provider whose density is logged, the GC's benchmark is the cheapest rate justification it will ever be handed.

LawFuel closes with a sentence we adopt as this memo's deadline rather than dispute: "Firms that begin the transition now will set the new market price. Firms that wait will be priced by it." What separates the two versions of the rising rate is not the rate. It is what sits beneath it.

5 · The mechanism · an operations layer beneath the lawyer

Here is the third branch, stated as a mechanism rather than a slogan. An operations layer runs beneath the lawyer: process engineering, tooling, verification workflows, the unglamorous compression of cost and cycle time on everything in the matter that is not judgment. Two things distinguish it from a software subscription. First, it delivers the density rather than promising it: the work the client receives in a billed hour is greater because the production beneath that hour was rebuilt, not because the lawyer typed faster. Second · and this is the load-bearing half · it measures the density: work units delivered, cycle time per task, logged per matter, so that what changed is a record rather than a renewal-season story.

Call the output the dense hour: a billed hour that carries more delivered work because cost and cycle are compressed and measured beneath the lawyer, so that its higher rate is an audit result, not a story. The measurement is not overhead on the mechanism; it is the mechanism's commercial point. The GC benchmark of 2027-28 audits hours per task and cost per outcome. The dense hour is the only version of the $1,200 rate that produces, on demand, the very evidence the benchmark exists to demand. The ops layer converts Section 4's threat into the rate's defense.

Notice what this does to the two-trade menu of Section 1. The earlier memo in this series stated its adoption law falsifiably: "adoption appears exactly where one of the three terms is removed, and nowhere else." The dense hour removes the hour's tax on the payoff · the rate rises with the density, so compression no longer cuts the top line · and adoption follows exactly where that memo predicted it would. Nobody adopted efficiency in thirty years because efficiency, decomposed, was the two losing trades. Everybody adopts a raise, because a raise is what the third branch pays.

Two boundaries, stated plainly. Who captures the savings between the firm and the operating entity beneath it · fee design, the dividend's address, the economics of the spread · is settled ground in "The AI Dividend Has an Address, and It Is Written in the Fee Clause" and "The Trust Spread Has a Published Closing Schedule," and this memo does not reopen either; here we price only the hour the lawyer sells the client. And one declaration of interest, owed before any table asks to be believed: Jopese operates exactly such a layer beneath a law practice, which means this memo describes its author's product. Judge the arithmetic, not the author; every figure below is modeled and the mechanism is stated so a hostile reader can audit it.

6 · The practice pushes back

Now the objection this memo must survive, from the 30-year managing partner, at full strength: "Your table is a unicorn. Rates are set by clients and reputation ladders built over decades, not by dashboards; no GC pays $1,200 because your ops layer has logs. Second, if my take-home doubles while the client's unit cost falls by half, someone is eating the difference, and your arithmetic smells like a consultant's free lunch. Third · and this is the one your table is hiding · those compressed hours you celebrate were not cost. They were profitable billings. The associate's hour cost me a salary share and collected a multiple of it; that spread is where firm profit actually lives, and your raise is funded by deleting my associate margin. And fourth, your own Section 4 kills you: once the GC's benchmark says the task takes four hours, they will not buy 1,400 hours at any rate. They will demand a fixed fee at their benchmark, and your dense hour dies alongside my plain one."

Concede what must be conceded, in order.

The rate ladder is not purchasable, and the table never claims it is. The $1,000-to-$1,200 move is a corridor entered over renewal cycles as the measured record accumulates, not an entitlement collected at go-live, and the assumptions block beneath the table says so in plain text. Logs do not set rates. They defend rate moves that the relationship has earned, in the one currency the 2027-28 buyer has announced it will accept.

The lunch is not free, and naming the payer is precisely where Villé's version fails. The surplus has a source: the cost and cycle time formerly priced into the matter as pyramid hours and process drag, now compressed beneath the lawyer at lower cost. The table shows the netting explicitly rather than asking you to trust it: collections rise $280,000 on the rate, the cost carried beneath the lawyer falls $120,000 on the compression, and those two movements · modeled, like every figure here · are the $400,000 raise. Nothing is conjured. Something is reallocated, which brings us to the real objection.

The leverage prong deserves the most honest answer in this memo, because the managing partner is right about the accounting. The compressed hours were not pure cost; a real fraction of them were profitable billings, and the spread on the leveraged hour is historically where partnership profit lives. Concede it without flinching: the dense hour does retire part of the associate-margin machine. Now ask what the alternative disposition of that margin is. Those leveraged hours · the routine, decomposable, benchmarkable middle of the matter · are exactly the hours the GC's internal benchmark prices first and most brutally, because they are the easiest to time. The margin the partner is defending is the renarrated hour's margin, and Section 4's calendar says it is the audit's first casualty, on the buyer's schedule and at the buyer's number. The choice on the table is not "keep the associate spread or take the dense hour's raise." It is "lose the spread to the benchmark at a discount you do not set, or convert it into a rate the benchmark certifies." The KPMG episode is what the first option looks like, executed by a client who measured before the provider did. The dense hour does not delete the leverage margin. The benchmark deletes it; the dense hour is the conversion that leaves a margin standing afterward.

And the fourth prong · the benchmark kills hour-buying entirely · is conceded as a possibility and survived, because the table is paper-agnostic. Suppose the GC wins completely and the matter reprices as a fixed fee at the benchmark. Run the same arithmetic under that paper: the lawyer's compensation per unit of delivered work and the client's falling cost per unit reconcile identically, because both are functions of the density, not of the billing format. The dense hour and the benchmarked fixed fee are the same economics wearing different paper. What dies in 2027-28 is the unmeasured rate, on any paper. What survives is whoever can show, per matter and per task, what an hour of theirs now carries.

7 · One hour, two ledgers · and the volume variant

The table below is the argument in one artifact, and it is the screenshot this memo expects to be forwarded without its prose. Assumptions, stated plainly: every figure in this table is a modeled illustration, not a report of any actual practice, client, or engagement. A single partner's book, 1,400 billed hours a year, held constant across both columns deliberately: nothing here asks anyone to work more. The density factor of 2.5x is modeled and deliberately conservative: LawFuel's own headline scenario, "From 40 Hours To 4," implies 10x on the narrow tasks it describes; 2.5x is a blended factor across a full matter mix, most of which is not the narrow task. The rate move of $1,000 to $1,200 is a corridor entered over renewal cycles as the measured record accumulates, not a day-one entitlement. The cost carried beneath the lawyer · pyramid hours, process drag, overhead allocation · is modeled at $1.00 million today, compressing toward $0.88 million under the ops layer. No fee-clause, savings-capture, or spread arithmetic appears anywhere in this table; that ledger belongs to the two memos cited in Section 5.

One Hour, Two Ledgers · modeled

The netting, explicitly: collections up $280K on the rate, cost beneath down $120K on the compression, equals the $400K raise. The partner's raise and the client's discount are the same row read from opposite sides; the ops layer is what lets both be true at once.

One variant completes the picture, because the rate channel is not the only way this adoption law executes. In February 2026, Fortune reported Lawhive's $60 million Series B and one number that belongs beside the table above: "Lawhive lawyers earn as much as 2.8 times what they would make at a traditional practice." The mechanism is volume, not rate: about 500 lawyers working through its platform across three regulated law firms, handling routine consumer matters largely on fixed fees, with AI letting each attorney carry far more cases than the 80 to 200 clients a consumer practice typically juggles. Read it in this memo's terms, which are ours and not Lawhive's: the same operations layer sits beneath the lawyer, the same density is delivered and measured, and the same thing is actually adopted · not efficiency, a raise. The rate channel densifies the hour for the partner whose market prices hours; the volume channel densifies the caseload for the lawyer whose market prices matters. Two channels, one law.

Thirty years of legal technology asked the lawyer to buy a proposition that decomposed, on the lawyer's own ledger, into hidden margin or self-funded fee competition, and the profession's refusal was arithmetic, not temperament. The third branch puts a different proposition on the desk: same hours, higher rate, a client whose cost per unit of delivered work falls by half, and a record underneath that survives the only audit the 2027-28 buyer intends to run. Nobody adopts efficiency. Everybody adopts a raise.

This memo is published by Jopese, a legal management services organization operated by HIRO PARTNERS LLC, a Texas limited liability company. It is offered for educational and analytical purposes only. It is not legal, tax, or investment advice, and it is not an offer to sell or a solicitation of an offer to buy any security or service. Jopese is not a law firm and does not provide legal advice or legal services; legal services are delivered by an independent law firm under a separate engagement in which Jopese does not participate. The "One Hour, Two Ledgers" table and every dollar figure, hour count, density factor, and percentage in it are hypothetical modeled illustrations and do not describe any actual firm, lawyer, client, agreement, or transaction, including any to which Jopese is a party. References to specific companies, publications, surveys, executives, and market developments are drawn from public sources and are provided as market commentary, not as an endorsement, a recommendation, or a representation of any relationship. "The dense hour" and "densification" are the authors' analytical labels, not industry terms of art.