L.E.K. Consulting's framework for how investors evaluate legal MSO opportunities publishes exactly two operating numbers: a midsize law firm generating $20 million in annual revenue pays its management services organization an $18 million annual fee, and the MSO's actual cost of delivering the services is $15 million. The MSO earns $3 million. The firm keeps $2 million. Now apply a hypothetical 30 percent AI cost shock to that published illustration: $15 million of cost becomes $10.5 million.
Under the flat fee L.E.K. describes, the MSO's revenue holds at $18 million and its EBITDA goes from $3 million to $7.5 million. Margin expands dollar for dollar with the savings.
Now restate the identical economics as if the same deal had been papered cost-plus at the implied 20 percent markup · this is our counterfactual restatement of L.E.K.'s flat-fee economics, not L.E.K.'s deal. Revenue is now a function of cost. Fifteen million dollars of cost carried an $18 million fee; $10.5 million of cost carries $12.6 million. The same automation that just enriched the flat-fee MSO has deleted $5.4 million of the cost-plus MSO's revenue and cut its EBITDA from $3 million to $2.1 million. That MSO spent real capital building AI so that its client could save $4.5 million, and then paid a $900,000 fine out of its own margin for the achievement.
Same company. Same cost base. Same AI. Opposite income statement. The difference is one sentence in the management services agreement.
1 · Same AI, opposite income statement
The instinct, when a deck says "AI-powered legal MSO," is to diligence the AI. The arithmetic above says the instinct is misaimed. The sign of AI's effect on an MSO's profit and loss statement · not the size, the sign · is set by the fee clause, before the technology is switched on. An automation program that is accretive under one fee regime is dilutive under another, with no change to the engineering.
There is a third regime on the menu, and it is the most instructive one. Suppose the fee were written as a percentage of the firm's revenue, set at the $18 million equivalent · a hypothetical regime, because as we will see it is also a prohibited one. The fee now depends on the firm's top line and not at all on the MSO's costs. The 30 percent shock leaves revenue at $18 million while costs fall to $10.5 million: EBITDA goes from $3 million to $7.5 million, identical to the flat fee on the cost side. But the percentage regime carries a kicker the flat fee lacks. The MSO is long the firm's growth, so every dollar of new firm revenue the AI enables flows partly to the MSO as well. Note what the alignment runs through: not a sharing of the savings, which the MSO keeps outright, but the firm's own top line. Incentives point the same direction on both sides of the contract.
That is the aligned trade. It is also the forbidden one. Texas Ethics Opinion 706, the most direct authority on point, holds that "a lawyer may not agree to pay a nonlawyer-owned vendor based on a percentage of the revenues of the lawyer or the lawyer's firm" · prohibited fee-splitting under Rule 5.04(a), a ban Holland & Knight reads as covering gross and net alike.
One footnote before the menu closes. A fourth design circulates in live deals: fees set per lawyer or per unit of usage, documented in Hunton Andrews Kurth's overview of law firm MSOs. Whatever their prevalence, their exposure is easy to state · they are long attorney headcount, which is precisely the variable AI shrinks, so MSO revenue falls through a different channel. A second efficiency fine, levied per seat instead of per dollar. We name the channel and stop; the macro version of that claim is outside this memo's scope.
2 · The efficiency fine
Here is the uncomfortable part. Rank the three regimes by ethics risk, the way the market actually ranks them, and you get: cost-plus and flat fee safe, revenue percentage radioactive. Sidley's March 2026 analysis of private equity investment in US law firms lists the lower-risk fee designs as "fixed fees for defined services, with scheduled increases," cost-plus arrangements built on "documented costs plus an agreed fixed margin," and "fair market value pricing, supported by third-party benchmarking." Holland & Knight's reading of Opinion 706 draws the same map from the regulatory side: the management fee "must generally be structured as a 'flat' monthly fee, 'cost-plus' fee or another fee structure that is not directly tied to the revenues" of the firm. Even the newest statute draws the same line: California's AB 931, signed in October 2025, bars California lawyers · for contracts entered into from 2026 through 2030 · from fee-sharing with out-of-state nonlawyer-owned structures while expressly exempting flat-fee arrangements that do not scale with recoveries. The verdict, across the opinion as commentators read it and now a statute, is consistent: fees keyed to the MSO's own costs or fixed in advance are the conservative designs; fees keyed to the firm's revenue are the violation.
Now rank the same three regimes by their economics under an AI cost shock, and the ordering exactly reverses. The percentage fee, banned, is the aligned trade. The flat fee, outside the ban, captures the savings but on a timer, as Section 5 will price. And cost-plus, the design with arguably the cleanest ethics pedigree because the markup is visibly tethered to documented cost, is the economically toxic one. Under cost-plus, every dollar the MSO automates away reduces its own revenue by a dollar plus the markup. The contract does not merely fail to reward efficiency. It fines it · twenty cents of the MSO's own profit, in our counterfactual, for every dollar of cost it engineers out. Call it what it is: the efficiency fine. A cost-plus MSO that deploys serious AI is running a buyback of its own income statement, and the "conservative" MSA turns out to be structurally anti-AI.
This is the inversion the market has not yet priced. The risk ranking investors carry into these deals is the pre-AI ranking, built when the only question was regulatory survival. AI flips the risk ordering of the fee menu, and the participants ranking cost-plus as the safe harbor are reading last cycle's chart.
3 · This is transfer pricing, and procurement law is the senior literature
Name the move precisely and the precedents appear. An MSO fee clause is transfer pricing: a price set between a law firm and its captive services entity, where the price allocates surplus between two related balance sheets. Legal is treating this as a novel ethics question. Two older literatures have already adjudicated the exact instrument, and both reached verdicts worth importing.
The first is federal procurement. The fee-as-percentage-of-cost design has a name there, the cost-plus-percentage-of-cost contract, and a status: banned. "The cost-plus-a-percentage-of-cost system of contracting shall not be used," reads FAR 16.102(c), a prohibition carried in statute at 10 U.S.C. § 3322(a) and 41 U.S.C. § 3905(a) and enforced for decades, reaching even subcontracts under non-fixed-price primes. The rationale is precisely the efficiency fine read from the buyer's side: when the fee grows with costs, the contractor is rewarded for inefficiency and punished for thrift. Federal procurement looked at the instrument legal ethics ranks as conservative and concluded it was the one contract form too perverse to sign. And procurement did not stop at the ban; it built the cure. The cost-plus-incentive-fee contract, FAR 16.405-1, shares cost savings between buyer and contractor on an agreed formula · government gain-share, the standard exit ramp from cost-plus perversity.
The second literature is the business-process-outsourcing market, which ran the experiment live when AI arrived. Efficio's analysis of financial-services outsourcing states the commercial reset plainly: "Full-time-equivalent pricing is increasingly misaligned with AI-enabled delivery. If a provider can process significantly higher volumes with materially fewer people, pricing must move accordingly." HFS Research has put numbers on the migration: on average, 30 percent of 2024 GenAI engagements already include outcome, consumption, or risk-sharing components, and providers expect that share to exceed 50 percent by 2026. Procurement banned the contract; the services market began abandoning it within a single pricing cycle of AI's arrival.
So every other market that met this misalignment had the same answer waiting: gain-share, outcome pricing, let the parties split what efficiency creates. Legal's version of that ramp runs straight into Rule 5.4.
4 · The forbidden alignment device
Read Rule 5.4 as economics rather than compliance and its effect on this menu is exact: it bans the proven alignment device. The percentage-of-revenue fee, the structure under which the MSO's automation incentives and the firm's growth incentives point the same way, is the one Texas Opinion 706 forecloses. What the rule leaves behind is not a binary menu, though. It is a binary plus an unpriced corridor: the savings-share clause keyed to the MSO's own cost base rather than the firm's revenue · procurement's CPIF formula transposed into legal · which no opinion has blessed and no opinion has condemned. That corridor being unpriced legal risk is itself the underwriting point. The one path to alignment carries a risk premium nobody can yet quote, because nobody has litigated it.
The healthcare contrast sharpens the irony. In healthcare MSO arrangements, percentage-of-revenue management fees draw scrutiny for the opposite fear: that they push utilization up. Nixon Peabody and Stout's June 2026 analysis of AI-enabled MSO services approaches this fact pattern from exactly that side · fair-market-value compliance, fee-splitting and corporate-practice risk, and the observation that AI intangibles do not fit traditional MSO valuation frameworks. It is a valuation-and-compliance treatment. The incentive side of the same fact pattern, which fee design makes the MSO want the AI to work, is the territory this memo is claiming, and as far as we can find, it is unclaimed.
The closest prior art deserves to be engaged by name. Reed Smith's analysis of what the AI capital gap means for firms considering MSO structures prints the right contrast: "Flat-fee arrangements place greater execution risk on the MSO... Cost-plus structures offer more predictable returns to the MSO but shift cost variability back to the firm." It is one inference from the whole point, and it does not draw the inference. It never states which party captures AI-driven savings under each model, and it never states that cost-plus mechanically shrinks MSO revenue as automation succeeds. Those two sentences are this memo.
5 · A dividend with a maturity date
If the percentage regime is prohibited and cost-plus is the efficiency fine, the investable regime today is the fixed fee. But be precise about what the fixed-fee MSO actually owns. It captures 100 percent of the AI dividend only until the fee resets. At renewal, or at any scheduled re-benchmarking, the firm's counsel arrives with the new cost reality in hand and negotiates the fee toward it. The dividend is not a perpetuity. It is a term annuity, and its duration equals the MSA's reset interval. Renewal mechanics convert directly into AI-thesis duration · the lens built in "The MSA Is the Cap Table," the first memo in this series, term as duration, applies to this clause with unusual force.
The credit market already prices MSAs this way, even if the equity narrative does not; the lender evidence · the underwriting of MSA term, renewal, and termination provisions, the collateral assignments of the agreement itself · is walked in full in "The MSA Is the Cap Table," and we will not re-run it here. The people whose money is senior have already concluded the fee clause is the asset. The question AI adds is how fast that asset amortizes.
Here the strongest objection deserves the floor. It runs: your fixed-fee dividend is an artifact of a stale benchmark. The fee must sit at fair market value to stay inside Rule 5.4, and after an industry-wide 30 percent cost shock, the fair market value of those services falls too. Competent firm counsel forces the fee down at the first reset; FMV scrutiny of the kind Nixon Peabody flags in healthcare compresses it from the regulatory side. The dividend always escheats to the firm at the speed of FMV re-benchmarking. Cost-plus is at least permanently safe, and the AI thesis lives in FMV, not in the fee clause.
Three answers, in ascending order of force. First, FMV is a range, not a point, and the benchmarks it is built from are peer comparables that themselves lag the shock; the dividend is the area between the falling cost curve and the lagging fee curve, a finite but real and underwritable asset. Second, "permanently safe" is the wrong reading of cost-plus: a contract that fines its holder for efficiency is not safe in a market where efficiency is arriving, it is merely slow. Third, and decisively: reset cadence, benchmark source, and re-benchmarking triggers are themselves negotiated clauses. The objection does not refute the thesis; it relocates it two paragraphs down the same document. If the dividend's duration is set by the reset mechanics, then the AI thesis lives in the fee mechanics, which is exactly the claim. We frame it as duration, not perpetuity, and we note that whether AI-era services economics ultimately re-rate at all remains contested. The fee clause decides who is positioned for the answer.
6 · Drafting in the corridor
The corridor Section 4 left unpriced is navigable, and each instrument in it has a hard ethics edge. What follows is design analysis, not legal advice; every one of these requires counsel against the relevant state's rules.
Scheduled fee resets are the honest version of the steelman: a fixed fee with stepped re-benchmarking functions as a clawback ratchet, amortizing the dividend back to the firm on a schedule both sides can price in advance. Sidley's "fixed fees... with scheduled increases" shows the drafting machinery is already standard; the AI era simply means the schedule can run in either direction. Savings-share clauses keyed to the MSO's own cost base are the gray instrument itself: not a percentage of firm revenue, and therefore not what Opinion 706 condemned, but unblessed by any authority we can find. A design option with a counsel caveat attached, never a cleared path. Per-matter and per-unit pricing imports the BPO destination model, pricing output rather than input · with its own care needed wherever unit counts shadow firm revenue too closely. And incentive structures keyed to nonlegal operating metrics · turnaround time, error rates, system uptime · bolt alignment on without touching the fee base at all.
One honesty obligation runs through all four. The market is visibly drafting toward alignment: L.E.K. itself observes that because "the MSO fee is paid before partner distributions, successful deals build alignment through fair market value fees, scope adjustment terms and rollover equity." Every one of those instruments, pushed far enough, quietly re-couples the MSO's return to the firm's success · which is the very coupling Rule 5.4 polices. The corridor is real, and so is the wall it runs along.
7 · Read the fee clause before the deck
The market is already asking this memo's question, and asking it inside the old ranking. When Law360 surveyed private equity's leap into the MSO "frontier", UCLA law professor Scott Cummings put the skeptic's case in one line: "The question is, What's in it for private equity if they're not able to tap into revenues?" The question carries the pre-AI map in its premise · that the only prize is revenue capture, and that everything the rules permit is consolation. The arithmetic above returns a concrete answer. What is in it for private equity is the AI dividend, and who receives that dividend is not a market outcome. It is a contracted term, set one level down, between the firm and its MSO. The AI dividend has an address, and the address is written in the fee clause.
Which collapses this entire memo into one diligence instruction. You cannot evaluate an "AI-powered legal MSO" from its technology narrative, because two MSOs with identical AI can be opposite trades. The diligence question is not "how good is the AI." It is "who is contracted to receive what the AI saves, and for how long."
The table below is the whole argument in one artifact. Assumptions, stated plainly: the model MSO is L.E.K.'s published illustration ($18 million annual fee, $15 million cost base, $3 million EBITDA); the AI shock is a hypothetical uniform 30 percent reduction in the MSO's cost base; the cost-plus column is our counterfactual restatement of those economics at the implied 20 percent markup, not L.E.K.'s deal; the percentage column is set at the $18 million equivalent and assumes flat firm revenue, before any growth kicker; all figures ignore taxes, transition costs, the capital cost of the AI itself, and any mid-term fee renegotiation.
Same MSO, Same AI, Three Fee Clauses
| Cost-plus 20% (counterfactual restatement) | Fixed fee (L.E.K.'s published deal) | % of firm revenue (at the $18M equivalent) | |
|---|---|---|---|
| Revenue before shock | $18.0M | $18.0M | $18.0M |
| Revenue after 30% AI cost shock | $12.6M | $18.0M | $18.0M · fee independent of MSO costs; rises if the firm grows |
| EBITDA before | $3.0M | $3.0M | $3.0M |
| EBITDA after | $2.1M | $7.5M | $7.5M |
| Who captures the $4.5M of savings | The firm · and the MSO pays a $900K efficiency fine | The MSO, until reset | The MSO keeps the savings outright and is long the firm's growth |
| Incentive verdict | Anti-AI | Pro-AI, decaying | Aligned |
| Regulatory verdict | Outside Op. 706's ban, per H&K's reading | Outside Op. 706's ban · FMV-capped | Prohibited (Tex. Op. 706) |
| Federal-procurement analog | Banned CPPC, FAR 16.102(c) | Firm-fixed-price | CPIF gain-share, FAR 16.405-1 |
| What happens at renewal | Markup renegotiated | Dividend resets toward the firm | Percentage renegotiated; growth exposure persists |
A fourth regime in the wild: per-lawyer and usage-based fees are long attorney headcount, which AI shrinks · MSO revenue falls through a different channel, a second efficiency fine. We name the channel and stop; the macro version of that claim is outside this memo's scope.
And the five questions to put to any fee clause before crediting any AI slide:
- Regime. Cost-plus, fixed, percentage, or per-unit · what is the fee actually a function of?
- Markup base. If cost-linked, which costs, documented how, and who audits the base?
- Reset trigger and cadence. What reopens the fee, on whose initiative, how often · this is the duration of the AI dividend.
- Savings-share. Is there any clause splitting efficiency gains, what is it keyed to, and has counsel opined on it in this state?
- Benchmark source. Who supplies the FMV comparables at re-benchmarking, and how fast do they decay after an industry-wide cost shock?
"The MSA Is the Cap Table," the first memo in this series, read the MSA as the security, the only ownable instrument Rule 5.4 leaves behind. This memo priced its most consequential clause. Around that pair, "Carriers Drew the Line" asks who actually polices the line the fee clause must never cross, "Privilege Is a Zoning Law for Data" locates the data asset the same contract owns, and "A Legal Transaction Has a Bill of Materials" itemizes the cost base these regimes fight over. The thread is the same throughout: in legal MSO investing, the contract is not the paperwork on the thesis. It is the thesis.
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. References to specific funds, firms, advisors, transactions, ethics opinions, statutes, and regulatory developments are drawn from public sources and are provided as market commentary, not as an endorsement, a recommendation, or a representation of any relationship. Fee structures discussed are analyzed as economic designs only; their permissibility is jurisdiction-specific and requires advice of qualified counsel.