Skin in the Stack

This memo was supposed to open with a price table. The table took one working day to build and failed more often than it succeeded. Harvey. CoCounsel. Lexis+ AI. Westlaw Precision AI. NetDocuments. iManage. Litify. Filevine. Lawmatics. As of June 11, 2026, not one of them prints a price where a buyer can read it before a sales call. The pairs that survived are in the exhibit below; the names that refused are the finding. The most heavily lawyered software market in the economy, selling to the most contract-literate profession in the economy, will not show that profession a number. Markets behave this way for exactly one reason, and an economist named the reason in 1970. The premium you are paying is not the price of the software. It is the asymmetry, invoiced.

The diagnosis is not ours, and crediting it promptly is the fastest way past it. Mike Whelan printed it a month ago: "Legaltech is a lemon-scented market, especially since the infusion of artificial intelligence," he wrote in May 2026, in a piece this memo will return to, because Whelan got the disease right and then prescribed, in our reading, the opposite of the cure.

Here is the claim, in the third paragraph where this series puts its claims. The legal-tech premium is not a markup waiting to be negotiated away by a tougher RFP, a better benchmark, or pooled buying power. It is the information asymmetry itself, priced and invoiced, and no quantity of information delivered to the current buyer kills it, because the current buyer · every current buyer, including the most sophisticated intermediary money can hire · is an agent with no P&L exposure to whether the tools perform. The premium dies on one condition only: a principal buys. An entity whose own profit absorbs every overpay and whose own operation runs on what it bought. The fix is not better software and it is not better information. It is a different buyer.

1 · A market that hides its prices is telling you what it is

Start with the refusals, because they are data, not inconvenience. Every name in the opening paragraph sells software to lawyers; every one of them, checked on June 11, 2026, routes the question "what does it cost" into a demo request, a configurator that ends in a sales call, or a quote form. This is not unusual for enterprise software generally · per OpenView data cited by Vertice, only about 17 percent of software vendors above $25,000 a year publish pricing · but legal tech practices it with unusual completeness, and the direction of travel is toward more opacity, not less: Clio, long the legal market's most transparent pricer, now displays only its $49 entry tier and places the upper tiers behind a sales quiz.

Hidden pricing has a standard business rationale: price discrimination, deal-size protection, competitive concealment. All true, and all beside the point, because the rationale describes what the seller gains and the structure describes what the buyer has already lost. A vendor hides its price when the price will not survive comparison, and a whole market hides its prices when the whole market knows its buyers cannot run the comparison anyway. Opacity at this scale is not marketing prudence. It is the lemon condition operating in public.

2 · The diagnosis is premise, not reveal

The economics here are fifty-six years old and need exactly one paragraph. George Akerlof's "The Market for 'Lemons'" (Quarterly Journal of Economics, 1970) showed that when sellers know product quality and buyers cannot verify it, price stops carrying information, good products cannot earn a premium for being good, and the market degrades: quality exits, suspicion prices in, and what survives is the premium a buyer pays for not being able to tell. The model was built on used cars. It runs without modification on legal software, where the buyer cannot verify quality before purchase and frequently cannot verify it after.

Nor is the application to software procurement new. Jan Devos and colleagues at Ghent documented a market for lemons in small-business IT procurement years before legal AI existed: firms that lack internal technical proficiency depend on external expertise to buy technology, the dependence is itself the asymmetry, and adverse selection follows. What law adds to the Devos picture is an irony with money in it: the buyer here is not a proficiency-poor machine shop. It is the contract-literate profession, and it is in the lemon market anyway, which tells you the condition was never about intelligence.

One more piece of hygiene, because the academic literature already has lawyers in a lemon market and we are obliged to say which side. The credence-goods literature, with Gillian Hadfield's "The Price of Law" (Michigan Law Review, 2000) as its legal anchor, casts the lawyer as the lemon market's seller: the client cannot judge the quality of legal advice even after receiving it. This memo's move is the inversion, stated explicitly so nobody mistakes it for reinvention: in the technology market, the lawyer is the lemon market's buyer, sitting on the unsighted side of the same asymmetry the profession is accustomed to occupying from above. The seller's chair and the buyer's chair are the same chair, turned around.

So Whelan's diagnosis is the premise, and we move to what he declined to conclude. If the premium is the asymmetry priced, it should be measurable. It is.

3 · The premium, measured

The method is a port. Take a lawyer-specific tool, name the job it does, find the horizontal product that does the same job for the rest of the economy, and put the two public list prices side by side. The multiple that falls out is the asymmetry, made legible. The exhibit below contains every pair that could be built from public list prices as of June 11, 2026, and a second panel of the pairs that could not, which is its own evidence.

The Same Job, Two Prices

Panel A · the pairs that could be built (public list prices, per user per month, fetched June 11, 2026)

¹ HubSpot's own page, as displayed June 11, 2026, presents Sales Hub Starter at $9 per seat per month billed annually and $15 per seat per month billed monthly. The multiple runs off the $9 annual list, because this exhibit's entire authority is that it compares list to list and the annual list is what HubSpot prints; against the $15 monthly rate the row would read 3.3x, and both numbers are here so nobody has to recompute it for us. ² Box renders its prices by script on its own page; the ~$20 Business figure is corroborated through Vendr and Capterra and flagged as such rather than passed off as a clean list price. ³ ChatGPT Business reflects OpenAI's April 2026 seat-price reduction; two-seat minimum. Claude Team also sells a $100 premium seat; the $20 standard seat is the like-for-like comparator and the premium seat is noted so nobody has to find it for us. ⁴ Docketwise bundles immigration forms, intake, CRM, and e-signature in one subscription, so the DocuSign pair compares a bundle to a component and is flagged as a functional-bundle comparison.

Panel B · the pairs that could not be built. Harvey. Thomson Reuters CoCounsel. Lexis+ AI and Protégé. Westlaw Precision AI. NetDocuments. iManage. Litify. Filevine. Lawmatics. ChatGPT Enterprise. No public list price as of June 11, 2026. And one arrival this exhibit watched happen: Clio, mid and top tiers. Between drafting and publication, the Essentials and Expand prices this table once quoted went behind a "Get Pricing" button, leaving only the $49 entry tier in print; by this exhibit's own rule, a price a buyer cannot read is a price this table cannot quote. The names are the data: in a lemon market, the price is the first thing to hide. Third-party estimates circulate for several of these names; estimates are not prices, and they are not in this table.

Read Panel A honestly, including the spread. The premium is not uniform. The intake and e-signature row prints the table's gentlest multiples, 2.3x and 2.6x, because e-signature is the rare job here a buyer can verify in an afternoon; document management runs 6.6x to 9.4x, where verification takes a migration. The premium widens exactly as verification gets harder, and it concentrates where the least-informed, least-repeat buyers shop · the solo and small-firm tier, buying once, comparing nothing, paying 5.4x at Clio's entry tier against HubSpot's annual list. That gradient is not an embarrassment to the thesis; it is Akerlof's signature. And the exhibit demonstrated the thesis on its own sources while we were still checking them: between drafting and publication, Clio's mid-tier prices went dark behind the very quiz §1 describes. The exhibit's own pricing decayed the way lemon-market pricing behaves, which is the thesis demonstrating itself on the memo's own sources. The AI assistant row is the steepest in the table at roughly 12.5x, and the market commentary has caught up to it: Antti Innanen, quoted in Artificial Lawyer this month, put it without anesthetic: "At the end of the day, most of these companies are middle men. They buy tokens and resell them at a premium." That premium, he added, "has to be justified somehow, often through convenience, workflow integration, proprietary data, or customer concerns about risk." Vendor-side surveys put numbers on the top of the range · Irys's April 2026 pricing landscape estimates Harvey at $1,000 to $1,200 per seat per month against a $20 horizontal seat · but those are secondary estimates about a vendor that will not print a number, which is why they live in this paragraph and not in the table.

And before a Global 200 reader files this exhibit under somebody else's problem: notice that the two panels are the same finding in two disguises. At the bottom of the market, where prices are public, the premium is visible as a multiple. At the top of the market, the premium is invisible for the simplest possible reason: the price itself is withheld, so there is no multiple to print. A condition that shows up as a number where numbers exist and as a refusal where they don't is not a solo-shop entry tax. It is the market's operating system.

4 · Every printed remedy repairs the information. None repairs the buyer.

Two remedies are already in print, and they fail at the same node.

Whelan's remedy is costly signaling: vendors should earn credibility through expensive, hard-to-fake commitments · hands-on sales, real education, genuine community · rather than "lemony claims." The second remedy is independent benchmarking, with Vals AI's Legal AI Report the serious version: third-party evaluation, standardized tasks, published results. Both are improvements. Both deliver better information to the buyer. And neither can work as a cure, because the failure was never informational. The buyer of legal technology inside a firm is, with rare exceptions, non-repeat (a practice buys a DMS once a decade), unaccountable (the committee that chose the tool has dissolved by the time the tool disappoints), and unexposed (the cost lands in overhead, spread across partners, attached to no one's number). Information cannot change the behavior of a buyer who bears no consequence for ignoring it. Hand that buyer a perfect benchmark and you have a better-informed version of the same indifference.

This is also where Jordan Bryan's piece belongs, engaged by name and subsumed rather than rebutted. In "Why can't $4.3B in legal AI outcompete $20/month for ChatGPT?" Bryan argues that legal AI's premium is bought as coverage · firms paying up to minimize the risk of being caught flat by disruption. Call it what it is, disruption insurance · the coinage is ours, not his. As description, correct. As explanation, it is not a rival to the lemon account; it is the lemon account experienced from the buyer's chair. A buyer who cannot verify quality buys reassurance instead, because reassurance is the only thing on the shelf he can evaluate. Insurance-buying is what the asymmetry feels like from the inside.

The mirror image deserves a name too. Ivy Grey, writing on the PerfectIt blog, argued that lawyers' refusal to buy is itself rational: the tech is lousy and overpriced, so not buying is the sensible verdict. We absorb the point whole. Rational non-purchase is not a counterexample to the lemon market; it is the lemon market's textbook output · Akerlof's market-thinning, the good tool exiting because the buyer cannot distinguish it from the bad one and therefore, correctly, buys neither. Overpaying and not-buying are the same equilibrium's two exits.

What the same premium does to the vendors who charge it · their compression schedule, their exposure to the labs · is settled ground in "The Trust Spread Has a Published Closing Schedule" and is cited here, not re-argued. This memo stays on the buyer's side of the invoice.

So the strongest objection to "change the buyer" is obvious, and it is not hypothetical: the market already changed the buyer, at scale, at the top, and the premium did not die. Engage it by name.

5 · Harbor already exists, and the premium did not die

Harbor, which says it works with more than 80 percent of the Global 200, runs a group purchasing program with over $3 billion in pooled buying power across the vendors law firms actually use. This is the sophisticated intermediary the remedy literature dreams about: professional procurement, aggregated demand, real category expertise, installed at the very top of the market. It is the closest thing legal tech has to an organized buy side. And the equilibrium this memo describes survived its arrival intact: prices stayed dark, multiples stayed wide where they are measurable, and the panel of vendors who refuse to print a number kept growing.

The reason is not that Harbor is bad at its job. The reason is what kind of buyer it is. Harbor advises and aggregates; it does not eat its own cooking. Its P&L does not absorb the overpay when a recommended tool underdelivers; its own operations do not run on the stack it negotiates; its compensation does not move when a member firm's shelfware ratio moves. It is an agent · a skilled one · and an agent buyer can move the price level: the discounts are real, and we will concede them again, with numbers' worth of respect, in the pushback section. What an agent cannot do is break the price structure, because the structure is not made of prices. It is made of the buyer's indifference, and aggregating ten thousand indifferent seats produces a larger indifferent buyer with a better discount.

State the boundary of the claim precisely, because overclaiming here would be its own lemony signal. We do not say no technology fiduciary exists anywhere; the claim is narrower and checkable: in legal technology, agents and GPOs exist; principals do not. Every buyer currently installed in this market, from the solo browsing Clio's entry tier to the intermediary negotiating for the Global 200, shares one property: somebody else's profit absorbs the mistake. The premium has never once faced a buyer without that property.

So describe the buyer it has never faced.

6 · The principal buyer

A principal buyer is an entity with three properties, each the negation of one failure from section 4. It is repeat: it procures the same categories continuously, across matters and years, so it accumulates the procurement memory a once-a-decade committee structurally cannot. It is accountable: the entity that chose the tool still exists, under the same contract, when the tool disappoints. And it is exposed: it buys under a fixed or benchmarked fee, so every overpay is its own margin lost, every shelfware seat its own write-off, and · the property no agent can replicate · its own delivery runs on what it bought. A platform that underperforms does not show up in a member-satisfaction survey. It shows up in the principal's own cycle times, its own error rates, its own P&L, this quarter.

That is what a management services organization is, on the buy side. The MSO purchases the practice's technology stack as its own asset, operates the practice's business layer on that stack, and collects a fee that does not grow when the stack gets more expensive. It is the first buyer in the legal stack whose buying and whose operating sit on the same P&L. Akerlof's buyer cannot verify quality before purchase; the principal buyer verifies it the only way that has ever actually worked, by metabolizing it · running its own operation on the tool and reading the result off its own books. The asymmetry does not survive a buyer who consumes what he buys, which is precisely why no vendor pricing page has ever had to face one.

Bryan's endgame deserves its direct answer here, because it is the real rival to "change the buyer." His prediction is that lawyers cut out the middlemen and go to the foundation models directly: the $20 seat wins. The instinct is exactly right · it is the principal instinct, the buyer reaching for skin in the game · but it arrives without the principal infrastructure. An individual lawyer self-serving on a $20 seat is still a non-repeat buyer with no procurement memory, no evaluation apparatus, and no entity to absorb and learn from the mistakes; and the $20 anchors the price of tokens, not the price of delivered, supervised legal work, which is the thing the firm actually buys. Bryan's endgame is a principal impulse trapped in an agent's body. Why lawyers rationally don't buy at all is "The Partner Without a Computer Was Reading His Balance Sheet Correctly" ground, cited as the adoption-side twin and not re-walked: that memo is why they don't buy; this one is why they buy badly when they do.

7 · The practice pushes back

The objection, from a veteran COO or a 30-year managing partner, at full strength:

"We are not the naive buyer in your fable. My legal-ops team runs structured RFPs, we read the independent benchmarks, and we joined a GPO and banked real discounts · seven figures of them. And look at your own cure with adult eyes: a principal buyer who owns the platform under a fixed fee has every incentive to under-buy · to keep its cheap stack, skip the upgrade, and pocket the difference. You have not deleted the agency problem. You have moved it inside the walls and pointed it at me. An agent who overcharges me is at least an agent I can fire at renewal; your principal is married to me."

Concede what must be conceded, because all three legs are real. The GPO discounts are real money, and a competent legal-ops team is nobody's lemon-market caricature. The under-buy incentive is real: it is the mirror-image agency risk, and pretending a structure deletes agency rather than relocating it would be selling, not analysis.

Now the answers, honestly. On the discounts: a discount moves the price level inside an intact equilibrium, and the top of this market will not even print the price the discount comes off of. Panel B is the GPO's own habitat, and it contains no numbers, by the vendors' choice; the GPO's negotiated terms are themselves confidential. A discount you cannot verify, inside a price you cannot see, is not the equilibrium breaking. It is the equilibrium intact, with better seats.

On under-buying: the risk is bounded by the same mechanism that creates the alignment. The principal's own operation runs on what it buys, so under-buying is not a quiet pocketing of the difference; it surfaces as the principal's own delivery failure · cycle times, error rates, the service levels the MSA contracts for and the firm reads every month. A landlord can skimp on a boiler he never touches; a principal buyer lives in the building. The residual · where captured efficiency goes after the purchase · is fee-clause machinery, and it belongs to "The AI Dividend Has an Address, and It Is Written in the Fee Clause", handed off in this line.

And on the sharpest blade, the marriage: conceded first · switching an MSO is heavier than swapping a vendor, and anyone who tells you otherwise is hiding the ball. But "married" is not the same as "trapped," because the MSA is a negotiated instrument with the divorce machinery drafted while both sides still like each other: a stated term, termination rights, transition-services obligations that survive notice, benchmarking clauses that mark the fee to market on a schedule, and data and platform portability written at signing rather than begged for at exit. The design of that instrument is "The MSA Is the Cap Table: How to Read a Legal MSO Like a Bond Indenture" ground and is cited rather than re-walked. Meanwhile, audit what firing the agent actually buys you: you fire him into the same market, and your next agent shops the same dark pricing with the same absent exposure. The renewal-date freedom the objection prizes is the freedom to re-enroll in the equilibrium.

The honest summary: a principal buyer does not guarantee omniscient taste, and this memo has not claimed it does. What the principal deletes is narrower and decisive · the one buyer property the lemon equilibrium cannot survive without. Indifference.

8 · The own-cooking test

Strip the memo to its instruction, in a form you can forward. When any intermediary · GPO, procurement consultant, managed-services pitch, platform vendor · promises to fix your technology spend, ask one question: whose P&L eats the overpay? If the answer is yours, you are talking to an agent, and an agent can get you a better price on the same asymmetry. If the answer is theirs · if their own operation runs on what they buy and their own margin absorbs the mistake · you are talking to a principal, and you have found the one buyer the premium cannot survive. An agent buyer recommends the restaurant. A principal buyer eats its own cooking.

Jopese operates a legal MSO, and this memo therefore analyzes the buyer structure it operates; readers should weight that disclosure as they see fit, alongside the exhibit, which does not care who built it.

The market told you everything at the top of this memo, in the table that failed to build. A market that hides its prices has told you what it is. Hire a buyer it cannot afford to lie to.

---

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. All prices in the exhibit are public list prices as displayed on vendor pricing pages on June 11, 2026, are subject to change, and may differ from negotiated terms; corroborated and bundled figures are flagged where they appear, and any vendor or tier whose price was not publicly displayed on that date, including prices withdrawn from display during this memo's preparation, appears in Panel B rather than in the priced rows. References to specific companies, products, publications, and commentary are drawn from public sources and are provided as market commentary, not as an endorsement, a recommendation, or a representation of any relationship. The "lemon market," "agent," and "principal" framings are analytical lenses, not characterizations of any specific entity or transaction.