Don't Make Your Lawyers AI-Native. Make Your Matters AI-Native.

The most repeated verdict in the AI-services discourse is also the most useful one, and it is a verdict against the firms most likely to be reading this. Foundation Capital named the prize in April 2024: software was always a tool the customer still had to wield, but "in the services business, responsibility for achieving the desired outcome sits with the company selling the service," and AI now lets a services company deliver that outcome at something closer to a software company's margin. General Catalyst put a balance sheet under the idea in August 2025, backing AI-enabled roll-ups that aim for "software-level margins" by owning and operating the very businesses they automate. The investor's distillation of both has hardened into a single line at the pitch table: do not bring me a services business that added AI. Bring me one that is AI-native.

Every word of that is right except the part nobody says out loud. "AI-native," as the phrase is actually used, is a synonym for built that way from zero · the greenfield company with no legacy process to unwind and no professional to re-convince. It is a definition written from the venture seat, and it quietly disqualifies every business that already owns the one thing the greenfield company most lacks: clients, judgment, and a book of trust. For a law firm · the hardest case, because here the disqualification is written into the rulebook by statute · the definition reads like a death sentence. Rebuild from zero, or be re-rated as a dying annuity. This memo argues that the death sentence is a category error, and that the error is not in the firms. It is in the unit of analysis.

1 · The firm is the wrong altitude. So is the lawyer.

Ask what, precisely, is supposed to "become AI-native," and the consensus answer is the firm. That altitude is wrong in both directions at once.

It is too high. A firm is not a thing that can be rebuilt; it is a settlement among partners, a book of relationships, a thirty-year deposit of judgment, and a pipeline of live matters that cannot be paused while the renovation happens. "Make the firm AI-native" is, operationally, "stop the firm and start a different one," which is why the only genuinely AI-native law companies are the ones that started yesterday with no clients. You cannot rebuild the plane in the air, and a firm is never on the ground.

And it is too low, the moment the consensus pivots to its fallback · make the lawyers AI-native, send them to prompt-engineering training, hire a director of innovation. This mistakes the depreciating asset for the durable one. The half-life of a specific AI skill is now measured in months: the prompt patterns, the model quirks, the tool-of-the-quarter are obsolete by the next release, and a partner drilled on this spring's interface has been handed capital that is already amortizing. What does not depreciate is the ability to see a matter as a system · to decompose it into components, locate the handoffs, and know which steps carry the firm's judgment and which merely carry its time. That is not an AI skill. It is systems thinking, and it has a career-length half-life precisely because it is indifferent to which machine sits underneath it.

Between the firm that is too big to convert and the lawyer who is the wrong thing to convert sits the unit everyone skips: the matter. A matter is bounded, it is repeatable in its components even when it is unique in its whole, and · unlike the firm · it can be made native without anyone's permission and without stopping anything. You do not make a firm AI-native, and you do not make a lawyer AI-native. You make a matter AI-native, and then another, and the firm becomes native as the residue of its matters changing one at a time. Every argument that follows is a consequence of moving the unit of analysis down to where the work actually is.

2 · The barbell is really a build-versus-rent decision

The "AI-native or nothing" verdict does not land evenly across firm sizes, and the shape of where it lands is the first clue that the consensus is solving the wrong problem. It is a barbell. The pressure is survivable at the two ends and crushing in the middle, and the reason is not size as such. It is whether the firm builds the AI-native capability or rents it.

The small firm cannot build · no capital, no engineering bench, no innovation budget · and so it is forced to rent, which means it is forced, accidentally, into the correct posture. The very large firm can build, and a few will, but the largest firms are also where a firm-wide rebuild is most impossible to execute: a behavior change across hundreds of lawyers, each with a vote and a book, against an incentive structure built to resist exactly that. The disciplined large firm therefore also chooses to rent at the edges, routing work to an operated capability rather than boiling its own ocean. Both ends converge on the same answer from opposite causes: poverty forces it on the small, discipline recommends it to the large.

The mid-size firm is the one with just enough resources to make the expensive mistake. It has the budget to attempt a firm-wide AI build and not nearly enough to finish one, so it funds a transformation office, runs pilots that never leave pilot, and discovers two years later that it has paid for the costs of building and the costs of renting and captured the benefits of neither. This is Michael Porter's "stuck in the middle" reproduced exactly, forty-five years later, in a new technology: the firm that commits to no clean strategy earns below-average returns on the attempt. The barbell is not a claim about headcount. It is a claim that the build path is a trap at every size, and that only the firms forced or disciplined off it escape.

3 · The pod is a strangler fig

If the unit is the matter and the posture is rent-not-build, the mechanism follows on its own, and it turns out to be a pattern software has had a name for since the early 2000s. When engineers face a legacy system too critical to switch off and too large to rewrite, they do not attempt the big-bang replacement that fails. They build the new system around the edges of the old one and route new traffic to it, function by function, until the new system has quietly grown over the old and the legacy can be retired. Martin Fowler called this the strangler fig, after the vine that grows around a host tree until it has become the tree.

The pod is the strangler fig applied to a law firm. A pod is a single matter-execution cell: an AI-native operations layer · run by a management services organization, not the firm · takes a category of matter and runs its backend end to end, while the lawyer sits on top as the judgment-and-control layer, receiving finished components and a sign-off queue. New matters of that type route into the pod. The old firm keeps running at full speed beside it, untouched and un-paused. Nothing is rebuilt; traffic simply starts arriving on the new rails, and the share of the firm's work running on those rails grows one matter at a time. There is no transformation project to fund, no firm-wide behavior change to mandate, and no moment at which the firm has to stop being itself. It strangle-figs its way to native.

The pod earns its keep on two ledgers at once. The first is immediate: the matters running through it are produced at the operations layer's cost and speed from day one, so the efficiency transition starts on the first matter rather than after a two-year program. The second ledger is the one that compounds, and it is the more important of the two. Every matter a pod runs deposits reusable context · the decomposition, the templates, the resolved components, the institutional exhaust of having done this kind of work once already · so the pod is faster and better on the next matter of its type. The pod is not merely efficient; it is an appreciating asset, and the firm that feeds it is buying down the cost of its own future work. (Two boundary notes for readers of this series. How much an individual matter's cost actually falls as automation rises is a question of dependency topology, worked in full in A Legal Transaction Has a Bill of Materials; this memo claims the entry mechanism, not the cost curve's shape. And the deeper reason a pod can hold a portfolio income statement that a single firm cannot is The Matter Never Repeats. The Component Always Does. · the pod is where that componentization is operationalized.)

4 · What the lawyer is actually trained to do

Move the unit to the matter and the upskilling question answers itself, against the consensus. Inside a pod the lawyer does not operate the AI · the operations layer does · so training the lawyer to operate AI is training for a job the model has been put in the building specifically to remove. What the lawyer does instead is the durable work: decompose the matter into its components, hold the control points where the firm's judgment actually lives, and supervise the assembly that arrives from below. That is systems thinking, and it is the exact skill that lets a lawyer work with an operated backend instead of competing with it.

This is why the curriculum the consensus reaches for · prompt engineering, tool enablement, a model-of-the-quarter rotation · is depreciating capex dressed as investment. The skill it builds is obsolete on the release cadence of the underlying model. The skill the pod actually requires, the ability to read a matter as a process and own its judgment nodes, is the one professional capability AI does not touch, because it is the residue left after everything mechanizable has been mechanized. Train for the residue, not for the machine. The firm that drills its partners on this season's interface is renovating a room it is about to demolish; the firm that teaches them to see the system keeps the only asset that survives the next model.

5 · The last time the profession externalized execution

There is an obvious objection sitting in the profession's own memory, and it deserves to be met at full strength: we have seen the externalize-the-back-office movie before, and it was called offshoring, and it transformed nothing. The legal-process-outsourcing wave was real and substantial · Pangea3, founded in 2004, was bought by Thomson Reuters in 2010 and sold on to EY in 2019; Mindcrest and others built the same business · and for all of it, the commodity work moved and the firm did not change. The objection is correct about offshoring. It is wrong that the MSO is the same movie, and the two differences are the whole argument.

The first difference is speed, and speed is not a detail in the AI era · it is the binding constraint. Standing up an offshore capability was a multi-year program: a captive or vendor relationship, a physical center, a hiring and training pipeline, and the slow accretion of trust before any real work would flow. The legal industry's own ethics machinery records the lag · the ABA cleared the path for outsourcing in Formal Opinion 08-451 (2008), under the rule governing a lawyer's responsibilities for nonlawyer assistance, and adoption built over the years that followed, not the weeks. An AI-native pod stands up on software infrastructure and a library of already-built skills, so its setup is measured in weeks, not years. That difference would be merely nice in a slower decade. In this one it is decisive, because a transformation program that takes years to land arrives obsolete: the firm that chooses the multi-year build is choosing to still be transitioning when the transition is over.

The second difference is control, and here the MSO does not merely improve on offshoring · it inverts it. Offshoring moved work away from the lawyer, across organizational and geographic distance, into a place the lawyer could not see. The professional cost of that was named by the ethics rulebook itself: Opinion 08-451 cautioned that outsourcing makes competent supervision harder when the work is performed far from the lawyer, across great distances and time-zone gaps, because oversight degrades with distance and opacity. An operated AI pod runs the other way. The work is instrumented · every step logged, every component inspectable, every output landing in a review queue · so the lawyer supervising a pod has more visibility into the work than a partner ever had over an offshore vendor or, frankly, over a junior associate down the hall. Software is observable; humans are opaque. The counterintuitive result is that externalizing execution to an instrumented MSO returns more control to the lawyer than keeping the work in-house with people ever did.

That inversion is not a nicety; it is what the current rulebook now demands. Technology competence is already a duty · Model Rule 1.1's comment requires a lawyer to keep abreast of "the benefits and risks associated with relevant technology" · and the ABA's first generative-AI opinion, Formal Opinion 512 (2024), makes supervision, competence, and informed client consent affirmative obligations a lawyer cannot discharge through a tool that is a black box. A pod whose every step is logged and reviewable is not just a nicer way to work; it is the only externalization that lets a lawyer actually satisfy the duty the rules now impose. Add the dimension the rulebook does not score but every honest professional feels · the personal integrity of staying genuinely in command of work that goes out under one's own name · and the instrumented pod is the first externalization that strengthens the lawyer's grip instead of loosening it. Offshoring traded control for cost over years. The MSO trades neither: it buys speed in weeks and hands back more control than the firm started with.

6 · The objection at full strength

The hardest pushback comes from two seats at once · the managing partner and the skeptical investor · and it is worth printing in full before answering.

"You have dressed incrementalism in a software metaphor. The strangler fig is a euphemism for 'go slowly,' and going slowly is exactly what loses in a winner-take-all technology shift: while your incumbent strangle-figs one matter type at a time, the venture-funded AI-native firm has rebuilt the whole stack and is taking the market. The investor is right and you are rationalizing. And from the partner's chair, the pod is the largest outsourcing decision the firm has ever made · the entire backbone of how matters get produced, handed to an outside operator · which means it carries every control and dependency risk offshoring ever did, only deeper. You are not de-risking the transition. You are betting the firm on a vendor and calling it a vine."

Concede the true parts, because they are load-bearing. Yes, the pod is a real strategic dependency, and yes, the from-zero competitor is moving fast. Now the answers.

On speed: the strangler fig is not the slow option, it is the fast-to-first-value option, and the two get confused only because the consensus measures the wrong clock. A firm-wide rebuild produces zero delivered value until it lands, years out, if it lands. A pod produces native-cost output on its first matter, this quarter, and compounds from there. Incremental in coverage is not incremental in time-to-value · it is the opposite. The big-bang rebuild is the slow option that merely looks bold.

On the dependency: the objection is right that the pod is a serious externalization and wrong that it inherits offshoring's risk profile, for the precise reason Section 5 gave · the pod is instrumented and the offshore center was opaque. A dependency you can see into, audit continuously, and supervise to the standard the rules require is a different object from a black box across twelve time zones, even though both are "outsourcing" on an org chart. The risk of externalization was never the externalization. It was the loss of sight, and the pod is built to restore exactly that.

And on the winner-take-all fear · the part the investor believes most · turn it around, because it is the close.

7 · The decision rule, and the reframe the investor is missing

Compress the memo into the choice a managing partner actually faces, and the one an investor actually misprices. The partner is choosing among a small set of transition strategies, and they are not equal:

The scorecard below records the authors' analytical judgments, not measured data; the comparison is illustrative.

The decision rule travels in one line: prefer the strategy whose unit of change is the matter, whose first value lands in weeks, and whose every step you can see · and be suspicious of any strategy that asks the whole firm to change before any value arrives.

Which returns us to the investor who opened the memo, and the bet the consensus is misreading. The "AI-native or nothing" reflex funds the greenfield firm because it looks clean: native stack, no legacy, no partners to convince. But strip both candidates to what they actually possess. The from-zero AI-native law company has the technology and no clients. The migrating incumbent has the clients · the trust, the relationships, the live matter flow · and rents the technology through a pod. Ask which half is harder to acquire. A native backend can be bought, built, or rented on a known timeline; trust and a book of matters cannot be bought at any price, on any timeline. The investor chasing only the greenfield is funding the team that solved the easier problem and calling it the safer bet. The genuinely mispriced opportunity is the incumbent that keeps the irreplaceable asset and rents the replaceable one · the firm that does not restart, but migrates. Whether the services-to-software re-rating proves durable is contested and unsettled. What is not contested is which asset is harder to manufacture, and the migrating incumbent is the only candidate that holds it.

So the verdict survives, inverted. The destination really is AI-native; the consensus had that right. It had the route wrong, and the unit wrong. A firm does not become native by rebuilding itself or retraining its lawyers. It becomes native one matter at a time, on rails an operated MSO already runs · which is the topology Jopese was built to occupy.

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 transition-strategy scorecard and its assessments are the authors' analytical judgments, and any operational illustration is hypothetical; neither describes any actual firm, client, engagement, or transaction, including any to which Jopese is a party. References to specific companies, investors, publications, ethics rules, 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. "Make your matters AI-native," "the pod," and "strangle-fig the firm" are the authors' analytical labels, not industry terms of art.