Re-Engineering the Law Firm
Why the Time-Based Legal Business Model Will Not Survive the Age of Intelligent Machines
The current legal revenue engine has not survived because it was good business, nor is it the ancient tradition lawyers imagine. It is a twentieth-century import from the factory floor that became the profession’s dominant pricing model only in the 1970s. Artificial intelligence is now overturning the arithmetic that made it work. This article argues that the firms which thrive in the coming decade will not be the ones that adopt AI fastest, but the ones that re-engineer the firm itself — its workflows, its project and process management, the way it composes its deliverables, and ultimately its business model. It also argues that the choice is no longer optional: a new class of alternative legal suppliers is already building on different economics, and they will make time-based billing not merely outdated, but unaccepted.
I. The Reckoning We Have Postponed
I came to law from engineering and manufacturing, and I have never been able to unsee the profession the way an engineer sees a production line. When I walked onto a factory floor, the first question was never how to make the workers move faster. It was how the line itself was designed — where the bottlenecks were, which steps added value and which only added cost, and what the whole system was actually built to produce. The slowest factories I ever saw were not staffed by lazy people. They were staffed by skilled people trapped inside a design that no one had revisited in years.
That is the legal profession in 2026. We are skilled people working inside a design no one has seriously revisited in decades. The leveraged structure of the modern firm — a small group of partners atop a much larger base of associates — took shape around the turn of the twentieth century. The way we monetize that structure, billing in tenths of an hour, is much younger: hourly billing did not become the profession’s default until the 1970s. Both are now colliding with a technology that breaks the assumptions each was built on.
Richard Susskind has spent four decades warning that this would not hold, and his warning has finally collided with a technology capable of proving him right.1 The pyramid endured through the rise of in-house departments, through alternative-fee experiments, through three waves of legal technology. Susskind’s thesis is no longer contrarian; it is becoming the consensus, and the trade press now states plainly that the model that survived all of that will not survive artificial intelligence.2
My argument in this article is narrow and, I hope, useful. It is not that AI will replace lawyers. It is that AI breaks the economic logic of the traditional firm, and that bolting the technology onto an unchanged structure is the single most dangerous thing a firm can do — because it produces the appearance of modernization while leaving the fatal arithmetic intact. The following lays out what I believe re-engineering actually requires: decomposing the work, building real workflows, importing genuine project and process management, rethinking how we compose deliverables, and confronting the alternative suppliers who are about to make the old pricing model indefensible.
II. The Architecture We Inherited
To re-engineer a system, you first have to understand why it was built the way it was. The traditional firm is not irrational. It is a remarkably efficient machine for a particular task: converting junior labor into partner profit without risk to the company. There are very few professions where new junior talent is a revenue engine out of the gate, rather than an investment that needs to be trained and mentored.
The mechanism is leverage. A partner who can supervise the work of several associates captures the margin between what those associates are billed out at and what they are paid. Stack enough associates beneath enough partners and you have a pyramid whose economics improve as it widens at the base.3 The billable hour is the connective tissue that makes the modern pyramid run, because it translates junior volume directly into revenue. The more hours the base records, the more the apex earns.
This structure is not merely a habit; it is encoded in the legal profession’s organizational form. As John Armour and Mari Sako have shown, the professional partnership — what they call the P2 form — is exquisitely well-adapted to delivering bespoke legal advice through leveraged human labor.4 Profit-sharing among partners, the up-or-out tournament for associates, and the absence of outside capital all fit together as a coherent system. The genius of the design is also its rigidity: every part assumes that the unit of value is human time, and that more complex work simply requires more of it.
It is worth being precise about how recent the hourly model actually is, because lawyers tend to imagine the billable hour as a timeless feature of the profession. It is nothing of the kind. For most of American legal history, lawyers charged flat fees, frequently guided by minimum-fee schedules published by state and local bar associations. The billable hour traces instead to Reginald Heber Smith, who around 1913–14 took charge of the Boston Legal Aid Society and, borrowing directly from Frederick Winslow Taylor’s scientific-management studies of factory efficiency, built a timesheet system to track each lawyer’s hours so he could clear far more cases on a meager budget.5
For decades, though, the timesheet remained an internal budgeting instrument, not a method of pricing. Hourly billing did not become the profession’s default until the early 1970s, and it was cemented when the Supreme Court’s 1975 decision in Goldfarb v. Virginia State Bar struck down bar minimum-fee schedules as antitrust violations and pushed firms toward hours-and-rate pricing.6 The implication is worth sitting with: the model now defended as sacred tradition is younger than many of the partners practicing under it, and it was, from its origin, a piece of industrial efficiency engineering grafted onto the practice of law.
Once hourly billing took hold, the model’s internal logic was sound. The cost of producing a contract, a memorandum, or a discovery review scaled more or less linearly with the hours invested. A firm that wanted to do more work hired more people. The hour was a defensible proxy for value because, in a world without machines that could read and write, time really was the binding constraint.
III. The Productivity Paradox
Then the binding constraint disappeared, and the model’s greatest strength became its central contradiction.
The contradiction is simple enough to state in a sentence: under the billable hour, efficiency destroys revenue. If a tool lets an associate complete in twenty minutes what used to take eight hours, the client’s bill collapses by more than ninety percent — and so does the firm’s income from that task. This is the productivity paradox, and it is not hypothetical. Goldman Sachs has estimated that as much as forty-four percent of legal tasks could be automated or augmented by AI,7 and the categories most exposed — research, document review, first-draft preparation — are precisely the work that fills the leveraged base of the pyramid.
The numbers from inside large firms are stark. In interviews with AmLaw 100 chief operating officers and partners, the Harvard Law School Center on the Legal Profession documented productivity gains that beggar the traditional model: in one high-volume litigation matter, an automated complaint-response system reduced associate time from roughly sixteen hours to a few minutes.8 The same study found that the billable hour still accounts for at least eighty percent of fee arrangements — which is exactly why dramatic productivity gains threaten, rather than enhance, firm revenue.
One of the partners interviewed framed the opportunity as an “80/20 inversion.” Historically, perhaps eighty percent of a lawyer’s time went to collecting and organizing information and only twenty percent to genuine strategic analysis; the aspiration is to flip those proportions.9 That inversion is the optimistic reading, and I share the optimism. But it only materializes for firms that change how they capture value. For a firm that keeps billing by the hour, the inversion is not an opportunity — it is a revenue cut, because the eighty percent it used to bill for has evaporated.
This is why the consultancy literature now speaks of re-engineering the operating model rather than adopting tools. As one analysis puts it, if drafting time falls by thirty to forty percent while a firm continues to bill hourly, revenue declines unless rates rise proportionally — an increasingly untenable proposition under client procurement scrutiny.10 The Thomson Reuters Institute found that eighty percent of law-firm respondents expect AI to fundamentally alter how they price, staff, and deliver legal work.11 The profession has, in other words, already conceded the point in surveys. What it has not done is act on it.
IV. Decomposition: The First Discipline
If the disease is a structure that monetizes time, the cure begins with a discipline that most lawyers have never been taught: decomposition.
Susskind’s most durable contribution is also his most practical. He argues that we should stop treating a legal matter as a single, indivisible lump of “lawyering” and instead break every matter into its constituent tasks — then ask, for each task, what the most efficient method and source of execution actually is.12 Some tasks demand a senior partner’s judgment. Some can be standardized into a template. Some can be handled by a junior lawyer, some by a paralegal, some by a subcontracted provider, and a growing share by a machine. The point is that these are different kinds of work that have always been billed as if they were the same kind of work.
To the engineer in me, this is simply the recognition that a matter is a process, not an event. You cannot improve a process you have never mapped. The firms that struggle with AI are almost always firms that have never decomposed their work and therefore have no idea which tasks are candidates for automation, which are candidates for delegation, and which are the irreducible core of professional judgment that justifies the engagement in the first place.
Decomposition is unglamorous. It feels like industrial engineering, not law, and that discomfort is exactly why the profession has resisted it. But every other concept in this article depends on it. You cannot build a workflow for a task you have not isolated. You cannot manage a process you cannot see. You cannot price work by value if you do not know what the work is made of. Decomposition is the foundation on which the re-engineered firm is built.
V. Incorporating Workflows
Once the work is decomposed, the next step is to stop performing each task as a one-off act of craftsmanship and start treating it as the output of a designed, repeatable workflow.
This is a genuine cultural rupture for lawyers, because we are trained to believe that every matter is unique and that judgment cannot be systematized. Some of it cannot. But far more of our work is repetitive than we like to admit, and the parts that repeat are precisely the parts that should be built into workflows rather than reinvented by hand each time.
I have lived this in my own practice. A large portion of transactional work consists of producing variations on a number of core instruments. For years, the profession’s answer to that reality was the form file — a drawer full of prior documents to be copied, pasted, and manually conformed, with all the transcription errors and version-control failures that implies. The re-engineered answer is different. It treats the document as the output of an assembly process: a master template with structured, conditional components; an intake step that captures the client’s facts once, cleanly, and flags internal inconsistencies before they propagate; and automated mechanics — cross-references, schedules, tables of contents — that update themselves rather than being rebuilt by a billing associate at midnight.
The lawyer’s role does not disappear in this model. It moves. The judgment goes into designing the workflow — deciding what the conditional logic should be, which provisions belong together, where the genuine decision points lie — and then into the substantive review of each output. The keystrokes go to the machine. That reallocation is the whole point: it is the 80/20 inversion made concrete at the level of a single deliverable.
The payoff is not only speed; it is expanded capacity and improved quality. One firm described how AI-enabled due diligence allowed it to review an entire document set where it had previously sampled only seventy-five percent, delivering a more thorough work product from the same resources.13 That is the workflow dividend. The firm does not simply do the old work faster and bill less; it does better work, and it does work it could not previously afford to do at all.
VI. Legal Project and Process Management
Workflows govern individual tasks. Managing a whole matter — and a whole firm full of matters — requires something the legal profession has historically treated as beneath it: project and process management as a genuine discipline, staffed by people who do it professionally.
In construction, manufacturing, and software, no one is surprised that complex work requires dedicated project managers and process engineers. The lawyer’s instinct has been the opposite: that managing the work is what the supervising attorney does in the cracks between doing the work. That instinct was affordable when the inefficiency was simply passed through to the client as billable hours. It is not affordable when clients are buying outcomes and predictability rather than time.
There is a cultural reason this discipline has been so slow to arrive, and the legal-industry analyst Jordan Furlong put it to me more sharply than I could put it myself: lawyers and engineers are about equally intelligent, but the unstructured minds tend to go into law and the structured minds into engineering.14 Lawyers, on this account, are inclined to see process and systems as guardrails that constrain their brilliance rather than as channels that focus and amplify it. Having spent a career on the engineering side before crossing over, I think the diagnosis is largely right — and it explains why so many firms reach for AI as a clever tool to hand to brilliant individuals rather than as the occasion to finally build the systems those individuals have always resisted.
It also explains a peculiarity that still astonishes me. The legal industry runs largely without the operational metrics any manufacturer would consider elementary — little systematic measurement of cost of goods sold against revenue, of cycle time, of rework, of the true profitability of a given matter or client. Hours billed has stood in for all of it. A firm that re-engineers around AI cannot operate that way; it has to measure what it produces and what it costs to produce. I suspect AI will finally force that reckoning, because you cannot manage — or honestly price — a delivery system you refuse to measure.
Legal project management imports a basic toolkit — scoping, budgeting, sequencing, risk identification, and disciplined communication — and applies it to the matter as a whole. Process management goes a layer deeper, asking how a recurring type of matter should be run every time, so that quality and cost stop depending on which associate happened to be assigned. These are not administrative afterthoughts. They are where consistency, predictability, and measurability come from, and those three qualities are exactly what sophisticated clients now demand.
Adoption has to be staged. The most useful framework I have seen describes a “crawl, walk, run” progression: lawyers first build a foundational understanding of what generative AI can and cannot do, then apply it to simple, practice-specific use cases, and only then take on sophisticated applications tied to defined business objectives — measured against concrete metrics such as time per matter, error rate, and write-offs.15 Transformation, as one technology leader put it, is not something you can drop on lawyers’ desks and expect to take hold overnight.
Consider litigation, where the discipline pays off across the entire lifecycle. A firm that builds a managed process can use AI at the intake stage to construct a case timeline, then carry that structured factual record forward so that discovery requests use language consistent with the complaint — reducing the vagueness and rework that plague unmanaged matters. The benefit is not a single clever tool; it is a process in which each stage feeds the next because someone designed it to.
This is also where new roles enter the firm. The re-engineered firm employs legal project managers, process designers, and knowledge engineers — people whose job is the system itself. To a traditional partnership, these look like overhead. They are not. They are the people who make the difference between a firm that uses AI and a firm that is built around it.
VII. Composing the Deliverable
Decomposition tells us what the work is made of. Workflows and process management tell us how to run it. The third question is what the work product itself becomes when intelligent machines are doing the first draft.
The deliverable stops being a thing a lawyer writes from a blank page and becomes a thing a lawyer composes and validates. The raw material — a first-draft contract, a research summary, an assembled estate plan, an initial discovery response — is generated quickly, from templates, clause libraries, and AI assistance. The lawyer’s contribution shifts from production to composition and judgment: selecting, arranging, correcting, and — above all — verifying.
Verification is not a footnote to this model; it is the core professional competency of the re-engineered firm, and the cost of skipping it is now a matter of public record. In Mata v. Avianca, lawyers filed a brief containing citations that an AI tool had simply invented, and the court sanctioned them for it.16 That case is the cautionary tale every lawyer now knows, and it teaches exactly the right lesson: the machine’s speed is worthless, even dangerous, without a human layer of judgment and quality control sitting on top of it. The right mental model is the one I use and suggest to every lawyer I consult with: treat the machine like a capable first-year associate — fast, confident, and occasionally, catastrophically wrong — whose work you would never file without reading it yourself. The firms that win will be the ones that build verification into the workflow as a required step.
There is a quiet irony here that lawyers should sit with. The skill that becomes most valuable is not faster drafting; it is the judgment to know when a draft is wrong. That judgment is the irreducible core that decomposition was always pointing toward — the twenty percent that becomes eighty. Composing a deliverable, in this new sense, means designing the components so that the assembled product is reliable, and then bringing genuine expertise to bear on the result. The lawyer who understands this is not threatened by the machine’s drafting ability. The lawyer is liberated by it and made more clearly accountable for the part only a lawyer can do.
VIII. The Alternative Suppliers at the Gates
Everything to this point describes how a traditional firm can re-engineer itself from the inside. The reason it must is that a new class of competitors is being built from the outside, on economics the traditional firm cannot match without changing what it is.
These are the alternative legal suppliers — not the law-adjacent outsourcing shops of a decade ago, but a wave of “full-stack” AI-native firms designed from the ground up with artificial intelligence at their core rather than retrofitted onto a partnership.17 The signal was unmistakable when Y Combinator, the startup incubator, explicitly challenged founders to start their own law firms, staff them with AI agents, and compete directly with incumbents. The ambition is not to sell software to law firms. It is to replace them.
The roster is already real. Garfield AI became, by its own account, the first fully AI-powered law firm authorized by the United Kingdom’s Solicitors Regulation Authority, focusing on small-claims debt recovery. Crosby pairs custom software with in-house lawyers for agentic contract review. Avantia runs a corporate practice with no billable hours and fixed-price services delivered through a proprietary platform, and was acquired by Carta. Tacit offers contract review at a fixed fee — starting around the price of a nice dinner — by having AI analyze a contract before a senior lawyer signs off. Eudia opened a regulated, AI-augmented legal-services business in Arizona for enterprise contract and diligence work.18 Manifest raised sixty million dollars at a valuation of three-quarters of a billion.19 These are not science projects. They are funded, regulated, operating businesses.
A vivid illustration arrived in June 2026, when three intellectual-property lawyers — two of them partners — left Goodwin Procter to launch Antheros, an AI-native IP firm for life-sciences companies.20 Their backgrounds rhyme with my own: a former pharmaceutical researcher, a former medical-device engineer, a Ph.D. scientist. They did not bolt AI onto an existing practice. In the managing partner’s account, they designed every workflow with AI as the base and then built hiring and fee structure around it. The firm maps each discrete IP task — gap analysis, prosecution, regulator correspondence — to the tool best suited to it, runs a standing “AI work team” to keep that mapping current as the tools change, and prices through subscriptions and flat fees rather than the clock. One founder’s explanation of why they built fresh rather than reforming an incumbent is the thesis of this article in a single sentence: they were unwilling to try to adapt an eighty-year-old structure to new tools.
The structural insight behind them was put well by one commentator who described traditional firm economics as a movie theater: the star partner sells the tickets, but the theater makes its money on the popcorn — the high-volume, industrial analytical work bundled in beneath the marquee name.21 Tech-first platforms may not have the marquee name, but they are engineered to capture the popcorn, and the popcorn is where the profit has always been.22
This is also why capital matters. The new entrants are financed differently. The legal-AI platform Harvey raised three hundred million dollars at a three-billion-dollar valuation; private equity is funding both the disruptors and, increasingly, the operations of incumbents willing to restructure.23 As one Stanford scholar observed, once you allow capital into the system and pair it with AI, you open the door to legal services being organized very differently from the traditional model of lawyers selling time inside a partnership.24 Capital plus AI is a different machine than partner profit plus billable hours, and it is being pointed directly at the most profitable, most automatable layer of the traditional firm.
Capital cannot reorganize a market the law forbids it to enter, which is why a quiet regulatory revolution matters as much as the technology. For nearly a century, Model Rule 5.4 barred nonlawyers from owning or investing in firms that practice law. That wall is coming down. In 2020 Arizona eliminated Rule 5.4 and, the following year, launched an Alternative Business Structure regime that lets nonlawyers hold economic interests and decision-making authority in a law firm; well over a hundred ABS firms are now licensed, and in 2025 the state approved KPMG Law US — the first Big Four accounting firm authorized to practice law in the United States.25,26 Utah and Washington are running parallel experiments; California, Florida, and Texas have so far held the line. But the dam has been breached, and capital now has a lawful channel into the practice of law.
For a practitioner in my own field, this is not an abstraction. The ABS structure is precisely what lets a multidisciplinary firm bundle estate planning, tax structuring, and wealth advisory into a single regulated offering, and it is what would let a bank, a multi-family office, or an accounting firm bring trust-and-estate work in-house rather than referring it out to a traditional firm.27 An estate-planning lawyer who assumes the work is safe because it requires a license should notice that the license is no longer the moat it once was. The moat now is judgment, relationship, and a delivery system the client cannot easily replicate — not the bare fact of admission to the bar.
What unites these entrants is that they never had to re-engineer anything. They were built system-centric from the first day, with no pyramid to dismantle, no partner-compensation formula to defend, and no installed base of billable-hour habits to unlearn. Where the traditional firm is a pyramid — wide at the base with junior labor — the AI-native firm looks more like an obelisk: a slender column of senior judgment sitting atop automated production, with far fewer juniors in between.28 That shape is the physical expression of the AI-intelligent operating model, and it is brutally difficult for an incumbent to copy without breaking the economics that currently pay its partners. Armour and Sako anticipated exactly this when they traced the migration from the traditional law firm toward the “next-generation law company” — an entity organized around technology and multidisciplinary inputs rather than leveraged human time.29
Jordan Furlong has framed the deepest version of this argument. Law firms and law schools, he observes, were built for a world in which legal intelligence — the capacity to analyze legal problems, draft legal instruments, and deliver legal solutions — was scarce, and that scarcity was the moat that let both institutions charge what they did.30 AI is converting that scarcity into abundance, and in the process it exposes how much of what firms sell has always sat uncomfortably close to commodity status: not the bespoke judgment of the best lawyers, but the leveraged, billable, done-the-same-way-everywhere work that fills the base of the pyramid. What follows is unbundling — the isolation of each function a firm performs so a client can ask which of them an AI now supplies adequately and affordably. The list, Furlong warns, is longer than firms would like to admit. The legal-technology analyst Nikki Shaver, whose work he builds on, puts the structural point bluntly: the billable hour is not merely how firms price, it is the atomic unit of their structure, finance, and culture, and removing it is like stripping a strand from the institution’s DNA.31
The squeeze is arriving from two directions at once. From below come the outcome-priced, “autopilot-native” services that take the lawyer out of the loop for narrow, well-defined bands of work and charge for the result rather than the seat. From above come the foundation-model providers themselves — Anthropic, OpenAI, and Google — now reaching directly into legal work; the 2026 launch of a dedicated legal offering from a frontier model provider, complete with practice-area plugins and connectors into legal software, signaled that the raw intelligence layer is climbing the value chain toward the firm.32
The competitive logic these suppliers introduce is unbundling. The most defensible, relationship-dependent, judgment-intensive work — governance advice, high-stakes litigation, complex negotiations where the law is genuinely unclear — will remain a moat around elite firms.33 But the industrial work that was bundled in beneath it is being pulled out and handed to platforms built on different economics. Crucially, those platforms are designed so that efficiency accrues to the client by default, rather than being quietly retained by the firm to protect its margins.34 That single design choice is what makes them so dangerous to the incumbent model.
IX. The End of the Billable Hour as Accepted Currency
Put the alternative suppliers together with the productivity paradox and you arrive at the conclusion that gives this article its urgency. The billable hour is not merely becoming outdated. It is becoming unaccepted.
“Outdated” would mean inefficient but tolerated. “Unaccepted” means clients will increasingly refuse it, because they will have a credible alternative that prices differently and a sophisticated reason to prefer it. We are nearly there.
Clients have grown far more sophisticated about how they buy legal services. Corporate legal departments now evaluate outside firms on their AI maturity — in many cases before those clients have even developed their own AI strategies — treating a firm’s operational capability as a selection criterion in its own right.35 And client patience with premium pricing untethered to value is thin: in recent market research, roughly one in four legal-services buyers reported never having experienced an outside firm that delivered excellent value despite premium pricing.36
Against that backdrop, the billable hour’s deepest flaw is exposed. It rewards exactly the behavior clients have decided to punish. It pays the firm more for taking longer and penalizes the firm for the efficiency the client now expects. A pricing model that is structurally misaligned with client interests survives only as long as clients have no alternative. The alternative suppliers remove that condition.
There is a deeper truth the billable hour always obscured, and it is worth saying plainly: clients were never really buying time. They were buying outcomes, risk reduction, clarity, and speed to a decision. Time was merely the proxy, the thing that happened to be easy to measure. Artificial intelligence removes the proxy and exposes the reality beneath it — that value is not labor but judgment applied to a problem, with accountability for the consequences. Once clients can see that, an invoice denominated in hours begins to look like a relic.
The replacement is not exotic. It is value-based and outcome-based pricing: fixed fees, efficiency guarantees, and above all transparency. As one analysis put it, the market will reward firms that price by outcome, guarantee efficiency, and are transparent, and the only real question is whether a given firm will lead that shift or be dragged into it.37 The temptation — and the trap — is to deploy AI quietly to protect existing margins while continuing to bill as before. That posture works precisely until a competitor passes the savings to the client by design.38 After that, it is indefensible.
X. The Re-Engineered Firm: Choosing a Form
If the time-based model is doomed and tools alone are not the answer, what does the re-engineered firm actually look like? The honest answer is that there is no single template — but there is a clear menu, and the fatal error is refusing to choose from it.
It helps to name the thing precisely. What I have been calling the re-engineered firm is, at bottom, an operating-model shift, not a branding exercise. Call it the AI-intelligent firm: a practice that has moved from a labor-centric model, in which output scales with bodies and hours, to a system-centric one, in which output scales with well-designed processes and is judged by cycle time, rework rate, predictability, and throughput. Both kinds of firm can employ excellent lawyers. The difference is how the firm is built to deliver work — what gets systematized, what gets measured, who does what, and what the client is actually buying. The forms below are simply different ways of housing that same shift.
In my opinion the deepest structural challenge is organizational form. Armour and Sako’s analysis is sobering for traditionalists: the new, AI-enabled business models require technological assets and multidisciplinary human inputs, and they are best served by centralized management — which sits in direct tension with the professional-partnership form that serves the old advisory model so well.39 When a firm tries to run both models inside one partnership, the complements conflict. Firms resolve that conflict either by contracting the new work out or by vertically integrating it into a separately managed entity. In plain terms: you may not be able to graft the new firm onto the old one without breaking one of them.
This is why the emerging incumbent responses cluster into three recognizable forms. The first is the lean, low-leverage boutique — a small group of senior, high-judgment lawyers whose capacity is amplified by AI rather than by armies of associates. The second is the hybrid firm, which keeps its traditional judgment practice but pairs it with a tech-enabled delivery arm, often capitalized separately, telling clients they can have the efficiency of a platform with the confidence of a known brand. The third is the legacy firm that looks much as it does today, using AI for internal efficiency while continuing to sell brand, global reach, and a single trusted relationship.40
The Thomson Reuters Institute frames the same choice as a set of business-model scenarios and warns against the one position that is not viable: the “dangerous middle” — the firm that is neither fully automated, nor elite and premium, nor large enough to compete on scale, nor protected by regulation.41 Such a firm does good work for clients at fair rates, and discovers that this is no longer enough to answer the question of why a client should keep retaining it. The firms that will dominate by the end of the decade are making their strategic choices now, while their competitors are still debating which tools to buy.42
Whatever form a firm chooses, the internal redesign rhymes across all of them: leverage shifts from junior labor to technology-amplified senior expertise; capital and infrastructure become things the firm invests in rather than distributes away; and talent is hired for new capabilities. Susskind has long argued that tomorrow’s lawyers will need engineering-style thinking, process design, and the ability to build and supervise systems of service delivery — skills no traditional law schools teach.43 The re-engineered firm is, in the end, a firm that has decided what it is for, and has rebuilt its structure to deliver exactly that and nothing it can no longer justify.
It is worth being explicit about what survives the unbundling, because that is what the re-engineered firm exists to sell. Furlong’s answer, which I share, is that the assets no machine can commoditize are a firm’s best people — those with exceptional judgment, credibility, and relationships — and the reputation that lets a client, a regulator, or a board rely on the firm’s advice and know the reliance was defensible even if the advice later proves wrong. What those assets share is trust. “Houses need load-bearing walls,” he writes; “societies need trust-bearing institutions.”44 The surviving firm is the one that keeps and deepens elite judgment and accountability while rebuilding its financial and cultural machinery to sell those things profitably — instead of selling the commodity hours that used to subsidize them.
XI. The Honest Counterweight
A practitioner writing for practitioners owes his readers the strongest version of the opposing case, and there is a real one.
The first caution is that incumbents are more durable than disruption narratives suggest. Big Law has navigated serious disruption before — the rise of large in-house departments, the alternative-fee movement, successive waves of technology — and the assumption that it will simply be displaced underestimates both the resilience of its model and its capacity to evolve.45 Many of the loudest AI announcements from large firms have produced more press than transformation, which cuts both ways: it shows complacency, but it also shows that the core franchise is strong enough to survive a great deal of complacency.
The second caution is regulation. In much of the United States, bar rules and licensing requirements mandate meaningful lawyer involvement in legal work, and some jurisdictions actively resist changes that would permit greater automation or non-lawyer ownership.46 Those rules are a genuine barrier, and they will slow the alternative suppliers in regulated practice areas even as the suppliers race ahead in commoditized ones.
The third caution is that the disruptors are unproven at scale. The tech-first platforms work in demonstrations and pilots; whether they deliver reliably across thousands of matters, and earn the trust incumbents spent decades building, remains to be seen.47 There is also a serious philosophical objection — that the most valuable legal work involves practical wisdom and contextual judgment that current AI handles poorly — and that objection has merit for exactly the judgment-intensive work that forms the elite moat.
There is a fourth caution, and it cuts against my own thesis hard enough that I am obliged to state it in full. Jordan Furlong — whose work runs through this article — doubts that traditional firms can be re-engineered at all. His view, which he put to me directly, is that the firm as we know it simply is not built to accommodate these changes and cannot realistically be rebuilt to accommodate them; the structured systems AI rewards are precisely what the traditional partnership was constructed to avoid. On that reading, the future of legal solutions will still feature lawyers in critically important roles, but it will not depend entirely on them and will not be led by them — it will be built and run, in significant part, by people who are not lawyers.48 The legal-AI platforms face their own version of the same trap. As one analysis of the sector’s “innovator’s dilemma” observes, the defensive move that would save an incumbent is the very move that destroys the business paying its bills today, because it means cannibalizing the high-margin seat to stand up a lower-margin services model that boards rarely approve. The pattern, that writer concludes, calls for a fundamentally different operating model — not a more polished version of the one firms run today.49
I take these points seriously. But notice what none of them denies: that the industrial layer is being automated, that capital is funding the automation, and that the dangerous middle is being squeezed. The counterweight does not rebut the thesis of this article. It refines it. The claim is not that every firm will be destroyed. It is that every firm must choose a defensible position, and that “keep billing by the hour and hope” is not one of them.
Furlong’s deeper challenge — and indeed a question that I have asked myself many times — can the existing firm structure be re-engineered at all, or will it have to be replaced — deserves more than a paragraph in response. It goes to the core of everything I have argued, and it is the underlying concept that has inspired this thesis and one I mean to answer in full.
XII. The Case That It Can Be Done
I have given the skeptics their due, and I want to be candid that some very capable people believe what I am about to argue is wrong. Their position is that the traditional firm is structurally incapable of reinventing itself, and that the future will be built only by new entrants. I understand why they hold it. I do not accept it. Re-engineering an established firm into a legal services company will be hard — harder than founding an AI-native one on a blank page — but hard is not the same as impossible, and the entire next decade of opportunity lives in the distance between those two words.
Start with history, because the “firms cannot change” thesis is refuted by the firms themselves. The leveraged pyramid was an invention, introduced around the turn of the twentieth century. The billable hour was an invention, imported from the factory floor and made dominant only in the 1970s. Each was a deliberate re-engineering of how legal work was structured and sold, and each was adopted across the profession within a generation. An institution that reinvented its core economic model once, well within living memory, cannot seriously claim that model is now a law of nature. The architecture we inherited was built; what is built can be rebuilt.
The strongest objection — that a single partnership cannot run the old model and the new one at once without the conflict destroying one of them — is real, but it assumes the firm must hold both models inside one undifferentiated entity. It need not. Armour and Sako’s own analysis points to the resolution: segregate the conflicting models into aligned but separately managed structures, joined by ownership or contract rather than forced into a single operating culture.50 The regulatory reforms now make exactly this lawful in a way it was not a decade ago. Under Arizona’s Alternative Business Structure regime — and Utah’s and Washington’s — an existing firm can stand up a separately capitalized, system-centric delivery entity, bring in the operators and outside capital it needs, and let its lawyer-owned practice concentrate on judgment work, all under common ownership.51 The very channel Furlong says the future requires — structures built and run in significant part by people who are not lawyers — can now be created by an existing firm, with its lawyers as co-owners rather than bystanders.
Notice, too, that the innovator’s-dilemma framework the skeptics invoke contains its own prescription. Christensen never concluded that incumbents are doomed; he concluded that they fail when they try to pursue a disruptive model inside the cost structure and metrics of the core business — and he prescribed the cure, which is to launch the disruptive effort in an autonomous unit with its own economics, insulated from the parent’s margins and measured on its own terms.52 A firm does not have to convert its mothership overnight, and it should not try. It can build the AI-native unit beside the existing practice, prove the model on a defined beachhead, and migrate work to it as the economics establish themselves. The dilemma is a warning about how incumbents fail; it is not a decree that they must.
The cannibalization trap is also far less binding than the BigLaw-centered debate assumes, because it bites hardest precisely where leverage and partner distributions run deepest. The large, broadly held partnership, dependent on leveraged-associate profit and annual distribution, genuinely struggles to vote for its own transformation. But most firms are not that firm. The small and mid-size practice has less leveraged labor to protect, simpler governance, and leadership that can simply decide. The barrier there is not capability; it is will — the willingness to trade some near-term profit for a defensible long-term position — and will is something a firm controls.
The cultural objection — that lawyers’ unstructured minds resist the very systems AI rewards — is real but not destiny. A re-engineered firm does not require every lawyer to become a process engineer. It requires the firm to hire the structured minds it has historically kept at the margins — the project managers, process designers, and technologists — and to concentrate lawyer judgment where only judgment will do. That is a staffing and design decision, available to any firm willing to make it. Re-engineering is not a single heroic leap but a sequence of provable steps: decompose one practice line, build the workflow, measure the result, capture the savings, and reinvest them in the next line. Done that way, the transition finances itself, and the firm becomes the thing its critics said it could never be — a legal services company — one matter type at a time.
None of this is easy, and I will not pretend the odds favor the average firm. Most firms will not do it. But “most will not” is a statement about courage and incentives, not about possibility — and it is precisely the gap the skeptics mistake for a wall. The firms that choose to re-engineer have a real path, and it runs straight through the disciplines this article has described. What it demands above all is the right people.
XIII. Talent: Hiring and Building for Judgment
Here again I need to acknowledge and rely on Furlong. At a recent conference we discussed his comments on new required legal competencies — a concept I had not spent much time thinking about until that moment, but was a missing piece to my own “how can this be accomplished” thought process. A re-engineered firm cannot run on traditionally trained lawyers doing traditionally defined jobs. If the work is decomposed, the workflows designed, and the deliverables composed and verified rather than hand-built, then the firm needs people who can actually do those things — and most of them were never taught how in law school.
The shift starts in the head before it shows up on an org chart. For a century, a lawyer’s value was measured by how much work he or she personally performed; more hours meant more value. The AI-intelligent firm inverts that. Value is now measured by the quality of the thinking a lawyer brings — how well they frame the problem, drive to a defensible answer, and design a process that delivers that quality consistently. The lawyer is moving from information gatekeeper, the person who held knowledge others could not reach, to judgment professional, the person who can be trusted with consequences. That is not a slogan; it is a different job description.
In my own teaching and practice I have come to rely on three baseline skills that matter more than fluency with any particular tool. The first is problem framing — the ability to turn a messy client story into the right question, to know which facts actually matter and which constraints are real rather than merely assumed. AI can generate fluent language, but it cannot reliably choose the right objective; that remains a human act. The second is verification discipline — the habit of treating speed as the enemy of care: checking sources, confirming jurisdiction, validating every citation, because persuasive is not the same as correct. The third is systems thinking, or workflow design — the recognition that the goal is never simply to “use AI” but to build a reliable process that produces consistent quality, with a clear line between what the tool may do and what only a lawyer should.53
I summarize it as a formula: relevance equals problem framing plus verification plus workflow design. Do those three things and AI is a force multiplier. Skip them and it becomes a risk multiplier, because it lets you produce confident-looking work faster without making it any more correct.
This creates a problem the profession has barely begun to confront. For generations, junior lawyers learned judgment by grinding through exactly the work that AI now absorbs — the document review, the first drafts, the research memos. Automate the apprenticeship and you risk a generation that never develops the judgment the senior roles require. The re-engineered firm cannot simply delete junior work and assume seasoned talent will materialize a decade later. It has to design new development paths on purpose — supervised verification, workflow design, structured exposure to real client decisions — that build judgment without the old volume of rote labor. This may be the single hardest part of the transition, and the firms that solve it will have an advantage that compounds for years.
The institutions upstream are beginning to respond. Law schools are experimenting with AI-integrated curricula and with tools that let students rehearse legal processes in simulation rather than only study appellate opinions, and the most forward-looking programs now treat technological fluency as a core professional competency rather than an elective.54 The firms that internalize this will hire differently, train differently, and ultimately think differently about what a lawyer is for.
The same scarcity logic reshapes legal education. As the law professor Michael Plaxton has argued, much of what law schools teach — the shared, interchangeable core delivered much the same way everywhere — is precisely what AI can now teach as well or better, one student at a time.55 What it cannot easily teach at scale is judgment, relationship, and professional formation. Furlong’s synthesis points the way: the school that survives becomes less a transmitter of scalable instruction than a customized facilitator of the formation that turns a graduate into a lawyer worth trusting. The firms that hire those graduates carry the parallel obligation — to complete the formation that the old apprenticeship of grind work once provided and that AI has now hollowed out.
XIV. Conclusion: The Decision Is Now
Susskind and his co-author once framed the future of every profession as a choice between two paths: a more efficient version of what we already do, or a genuine transformation in how expertise is delivered.56 The legal profession has spent a decade pretending it could stay on the first path indefinitely. It cannot. The arithmetic that sustained the time-based firm has been overturned, and a class of suppliers built on the new arithmetic is already at the gates.
The encouraging part is that none of this requires a leap of faith about distant technology. It requires disciplines we can adopt today: decompose the work, build real workflows, manage matters as designed processes, compose and verify deliverables rather than hand-craft them, and choose a business model honest enough to price by value instead of by time.57 The firms that do this will not be diminished by AI. They will be the ones clients trust to wield it.58 Put it the way I put it to every audience I speak to: the winners will not be the firms that say they use AI. They will be the firms that can say they built a disciplined, AI-enabled system that delivers better outcomes, faster, with accountability where it belongs.
I will end where I began, on the factory floor — fitting, since the billable hour was itself born there, in Taylor’s stopwatch studies before Reginald Heber Smith ever carried them into law. The first lesson every engineer learns is that when the inputs to a system change, you do not respond by running the old line faster. You redesign the line. The inputs to legal practice have changed more profoundly in the last three years than in the previous hundred. We can spend this decade running the old line faster and calling it innovation, or we can do the harder and more rewarding work of re-engineering the firm for the world we actually practice in. The clients have already decided which they will pay for. The only open question is which lawyers will decide in time.
Richard Susskind, Tomorrow’s Lawyers: An Introduction to Your Future (3d ed., Oxford Univ. Press 2023).
AI's Rise May Motivate Law Firms to Quit Their Traditional Ways, Bloomberg Law (Nov. 27, 2023).
John Armour & Mari Sako, AI-enabled business models in legal services: from traditional law firms to next-generation law companies?, 7 J. of Professions & Org. 27 (2020).
On the origins of the billable hour, see How Law Firms Ended Up With the Billable Hour Model, Thomson Reuters Institute (Feb. 2025); WilmerHale, Slice of History: Reginald Heber Smith and the Birth of the Billable Hour (Aug. 9, 2010); see also Frederick Winslow Taylor, The Principles of Scientific Management (1911); Reginald Heber Smith, Justice and the Poor (1919).
Goldman Sachs Global Investment Research (2023) estimated that as much as 44% of legal tasks could be automated or augmented by AI; the figure is now widely cited across the legal-industry literature.
Harvard Law Sch. Ctr. on the Legal Profession, The Impact of Artificial Intelligence on Law Firms' Business Models (Feb. 2025) (reporting qualitative interviews with AmLaw 100 chief operating officers and partners).
Id. (quoting Robert J. Couture on the prospect of an "80/20 inversion" in how attorney time is allocated).
P&C Global, supra note 3.
Susskind, supra note 1 (describing the decomposition of legal work into discrete, separately sourced tasks).
Strategic Law Firm Innovation in the Age of AI, Thomson Reuters (Dec. 2025) (panel remarks of Valerie McConnell, Jacob Edwards, and Kristina Bakardjiev).
Jordan Furlong, in correspondence with the author (June 2026); see also Jordan Furlong, The Unbundling of Lawyer Institutions, Law21 (June 17, 2026).
Strategic Law Firm Innovation, supra note 13.
Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023) (imposing sanctions where counsel filed a brief containing fabricated, AI-generated citations).
Lupl, 10 AI Law Firms to Watch in 2026 (May 2026).
Why 3 Attorneys Left Goodwin to Start an 'AI-Native' IP Firm, Law.com / Legaltech News (June 8, 2026).
Eric Greenberg, AI-Native Firms, Built by Private Equity, Will Strain Legacy Model, Bloomberg Law (Mar. 2026).
David Freeman Engstrom, quoted in Law, Disrupted, Stanford Lawyer (2026).
On Arizona's Alternative Business Structure program, see Ariz. Sup. Ct. order (2020) (eliminating Rule 5.4); Ariz. Code Jud. Admin. § 7-209; Arizona Courts, Alternative Business Structures: Questions & Answers, azcourts.gov; Arizona Alternative Business Structures: What KPMG Law US Could Mean for the Future of the Legal Community, Ariz. St. L.J. (Feb. 21, 2025).
Justin Henry & Roy Strom, KPMG Wins Approval to Launch First US Law Firm for Big Four, Bloomberg Law (Feb. 27, 2025).
KPMG’s ABS License Application: A Game Changer for Trusts, Estates, and Family Businesses?, Family Wealth Report (Feb. 2025).
Lupl, supra note 17 (describing the shift from the professional-services "pyramid" toward a leaner "obelisk" model, per Mary O'Carroll).
Armour & Sako, supra note 4.
Nikki Shaver, Law Firms Want to Change; They Just Can't, LegalTech Hub (2026), discussed in Furlong, supra note 14.
AI-Enabled Firms' Unbundled Offerings to Split the Legal Market, Bloomberg Law (Mar. 2026).
AI-Enabled Firms' Unbundled Offerings, supra note 33.
The New Economics of AI-Powered Legal Services, Thomson Reuters (Mar. 2026); see also 2026 Report on the State of the U.S. Legal Market.
The New Economics of AI-Powered Legal Services, supra note 36.
How Law Firms Can Lead the Agentic AI Era — And What Clients Now Expect, Harvard Law Sch. Forum on Corp. Governance (Mar. 2026).
Armour & Sako, supra note 4 (analyzing the professional-partnership, or P2, form and the centralized management that new business models require).
AI-Enabled Firms' Unbundled Offerings, supra note 33.
Thomson Reuters Institute, supra note 35 (describing the "dangerous middle" of firms without a clear strategic position).
Thomson Reuters Institute, supra note 35.
AI-Enabled Firms' Unbundled Offerings, supra note 33.
Thomson Reuters Institute, supra note 35.
AI-Enabled Firms’ Unbundled Offerings, supra note 33.
Furlong, supra note 14; the point was also made to the author directly in correspondence (June 2026).
The Innovator's Dilemma, supra note 32.
Armour & Sako, supra note 4.
See supra note 25 (Arizona, Utah, and Washington reforms permitting nonlawyer ownership and outside capital).
Clayton M. Christensen & Michael E. Raynor, The Innovator's Solution (Harvard Bus. Sch. Press 2003) (prescribing an autonomous unit with its own cost structure to pursue disruptive innovation); cf. The Innovator's Dilemma, supra note 32.
The three-skill framework and the "force multiplier / risk multiplier" formulation are the author's, developed for continuing-legal-education instruction; they track Susskind's call for new lawyer competencies. Susskind, supra note 1.
On law schools integrating AI into legal education, see Law, Disrupted, supra note 24; see also Susskind, supra note 1 (on the new skills and new legal roles the next generation will need).
Michael Plaxton, To Our Next Law Dean (2026), discussed in Furlong, supra note 14.
Richard Susskind & Daniel Susskind, The Future of the Professions: How Technology Will Transform the Work of Human Experts (Oxford Univ. Press 2015; updated 2022).
Richard Susskind, How to Think About AI: A Guide for the Perplexed (Oxford Univ. Press 2025).
How Law Firms Can Lead the Agentic AI Era, supra note 38.