In 2016 a prominent computer scientist said we should stop training radiologists, because deep learning would replace them within five years. Ten years later there’s a radiologist shortage. People said the same thing about lawyers, in the same tone, with the same confidence, and they were wrong the same way.
The technology didn’t fall short. Image classification got extremely good. The forecast missed because it treated a profession like a task, and a profession isn’t a task. It’s a bundle of tasks wrapped in licensure, liability, institutional trust, and the requirement that somebody you can name answers for it when it goes badly. The technology dissolved the task and left the wrapper standing.
Any prediction about law that ignores the wrapper will miss the same way. So here are mine, each with what I think drives it and — the part legal futurism always leaves out — the specific thing that would prove me wrong. Check back and score me.
Two clocks
One split explains most of it.
The cost of producing legal work is falling fast. The cost of everything gated by rules, courts, insurers, and trust isn’t.
Drafting, research, review, summarizing, extraction — the production work — is getting cheaper by an order of magnitude over a decade. Meanwhile the requirement that a licensed human sign, that a court accept a filing, that an insurer underwrite the risk, that a client trust the advisor, moves at institutional speed. Which is to say, roughly not at all.
Almost everything below comes from those two clocks running at different speeds.
Prediction 1 — The billable hour survives, and its share falls sharply
By 2035, hourly billing is still the dominant model for genuinely unpredictable work — litigation of uncertain scope, crisis work, complex transactions — while defined-scope work has mostly moved to fixed fees.
Why. The hour is a stand-in for value, and it only works when nobody can estimate the work. AI goes after estimability first: the more predictable a task gets, the sillier it looks to price it by the clock. Add the ethics piece — the ABA’s Formal Opinion 512 is explicit that if a task takes less time, you bill less time — and hourly billing on compressible work turns into a shrinking revenue line you’re professionally obligated to keep shrinking. Fixed fees get you out of that bind by letting the firm keep the efficiency it created.
How I’d know I’m wrong. Hourly billing holding steady or growing as a share of small-firm and consumer-facing revenue through 2030. If firms just keep billing the same hours for less work, I’m wrong that the constraint bites — and the access argument in this whole section goes down with it.
Prediction 2 — Document review as a staffed job mostly disappears
By 2032, linear human review of large productions is rare. The reviewing lawyer’s job becomes designing, validating, and sampling what a system produces.
Why. This is the most automatable high-volume task in litigation, and it’s already halfway gone. Courts have accepted technology-assisted review for over a decade, and language models are dramatically better at it than the classifiers TAR was built on. What’s left isn’t the reading. It’s defending the process: sampling protocols, recall estimates, privilege QC, and a lawyer who’ll certify how it was done.
How I’d know I’m wrong. Courts rejecting model-assisted review protocols, or a high-profile privilege waiver traced back to automated review that makes the practice uninsurable. Either one would freeze this for a decade.
Prediction 3 — Malpractice insurers regulate AI use before the bars do
By 2029, professional liability applications routinely ask detailed questions about AI verification workflows, and premiums differ based on the answers.
Of all twelve, this is the least obvious and the one I’d bet hardest on.
Why. Insurers move at underwriting speed. Bars move at rulemaking speed, which is years slower and needs consensus. Insurers already ask about cyber practices, calendaring systems, and conflict-checking procedures, and they price them. They’ve got every reason to ask “do you have a citation verification process” long before any bar requires one — and unlike a bar, they can act alone, right away, with a premium instead of a disciplinary file.
So the rules that actually govern AI in practice show up as a question on a renewal form. Firms that can document a verification gate get a better rate. That’s faster than discipline and better aimed, because it gets to you before the mistake instead of after.
How I’d know I’m wrong. Carriers treating AI as a general practice risk, with no workflow-specific underwriting, through 2030 — pricing it like weather instead of like something you can control.
Prediction 4 — Courts shift from “did you use AI” to “did you verify”
By 2030, most standing orders requiring AI disclosure have been replaced by rules requiring certification of verification.
Why. After the fabricated-citation cases, the first wave of judicial responses demanded that lawyers disclose AI use. That doesn’t survive contact with reality, for the obvious reason that pretty soon it’ll describe every document filed in every court — the same way “did you use a computer” stopped being a question worth asking. What courts actually care about is already covered by rules on the books: Rule 11 requires a certification “formed after an inquiry reasonable under the circumstances,” and that language doesn’t need amending to reach a lawyer who filed a brief they didn’t check.
Expect the orders to settle on asking what was verified and how, not what tool you used.
How I’d know I’m wrong. A durable, widely adopted disclosure regime — a model rule requiring affirmative AI disclosure that most jurisdictions adopt and keep past 2030.
Prediction 5 — The states drift apart before they come back together
By 2032, a handful more states permit some form of entity regulation or non-lawyer ownership; the large majority don’t; and the friction between states becomes the central regulatory fight.
Why. Arizona eliminated Rule 5.4’s ownership restrictions and licenses alternative business structures. Utah is running a sandbox on a fixed clock. Stanford’s five-year data shows the entities serving individual consumers and small businesses rather than corporate clients — which is exactly the population the access gap describes. That evidence will pull some states in. Most won’t move, because the profession in most states doesn’t want it.
The real pressure isn’t inside any one state. It’s between them. An Arizona ABS delivering services to a Minnesota resident over the internet raises a question nobody has a good answer to, and that question — not the merits of ownership rules — is what forces the issue.
How I’d know I’m wrong. Fast convergence in either direction: a wave of adoptions after a favorable multi-state study, or Arizona and Utah pulling back and the experiment getting written off.
Prediction 6 — The paraprofessional tier grows quietly and works
By 2033, licensed-paraprofessional programs exist in a substantial minority of states and handle a meaningful share of routine family, housing, and debt matters.
Why. This is the least glamorous of my predictions and the most likely, because it’s already happening and AI changes its economics in a big way. The old objection to limited-license providers was competence at the margins — a non-lawyer running into an unusual issue. A paraprofessional with good AI support and a clear path to escalate is a very different thing from one working off a checklist.
Minnesota’s program is the template: it started as a pilot and was made permanent effective January 1, 2025, with supervision by a licensed attorney and a defined set of case types. Narrow, built on evidence, boring, and built to last.
How I’d know I’m wrong. A repeat of Washington’s LLLT outcome at scale — programs launched and then sunset for low uptake. And watch for the trap I’ve written about before: these get judged against the ideal of full representation instead of against the actual alternative, which is nobody.
Prediction 7 — The bottom of the market gets served by something that isn’t a law firm
By 2032, a significant share of routine consumer legal need is met by products, not practitioners — and the fight is about unauthorized practice, not quality.
Why. Ninety-two percent of the civil legal problems of low-income Americans get no adequate help right now. Nobody’s serving that market badly. Nobody’s serving it. When the cost of producing competent guidance on a standard eviction defense gets close to zero, something will fill that hole, and what’s holding it back isn’t capability. It’s UPL statutes written back when the alternative to a lawyer was a stranger with a form book.
Watch how the coming argument gets framed: consumer protection. The real question is whether an imperfect product is worse than nothing, and for the 92% the answer’s usually no.
How I’d know I’m wrong. Aggressive, successful UPL enforcement against AI legal products, or consumer-harm evidence strong enough to justify it.
Prediction 8 — Leverage gets squeezed, and small firms gain
By 2033, the associate-leverage model that funds large firms is materially weaker, and the cost advantage of small firms over large ones narrows or reverses.
Why. Big-firm economics rest on billing junior time at a multiple of what it costs, on work that’s now the most compressible in the building — document review, first drafts, research memos, due diligence. Meanwhile a three-lawyer firm can get research and drafting capacity that used to take an institution. The gap in capability per lawyer shrinks. The gap in overhead per lawyer doesn’t.
This doesn’t kill big firms. Bet-the-company litigation, cross-border transactions, and regulatory work with real institutional depth aren’t going anywhere. It squeezes the middle, and the middle is where most of the associates are.
How I’d know I’m wrong. Large-firm headcount and leverage ratios holding steady through 2032, or firms successfully repricing junior work as supervision instead of production.
Prediction 9 — Verification becomes a named, billable job
By 2030, “verification” is a recognized role in litigation practice the way document review was in 2005.
Why. If producing work is cheap and certifying it isn’t, the value piles up in certification. Somebody has to establish that the authority exists, says what it’s cited for, and remains good law; that the extracted date is the right date; that the summary didn’t invent an admission. That work doesn’t compress the way drafting does, because it’s the check on the compressed output.
Watch what that does to status. Verification today is invisible, unglamorous, and mostly done by whoever’s most junior. Once it’s the scarce input, it stops being junior work.
How I’d know I’m wrong. Verification getting swallowed entirely by tooling — automated checking good enough that no human job forms around it. Possible for existence and quotation. I don’t buy it for “does this case stand for this proposition.”
Prediction 10 — Privilege and work product get litigated hard, and the answers are messier than anybody expects
By 2029, there’s meaningful appellate authority on whether AI prompts, retrieved context, and model outputs are discoverable in litigation.
Why. Every one of these systems leaves stuff behind: the prompt, the retrieved documents, the intermediate draft, the transcript. Is that attorney work product? A privileged communication? Does a litigation hold reach it? Does sharing a client’s facts with a vendor waive anything?
These questions have plausible answers under existing doctrine and only a handful of trial-court orders, which have come out different ways. That’s exactly the setup for expensive discovery fights. My weak prediction: the answers land roughly where work-product doctrine already sits — protected as reflecting counsel’s mental impressions, with the usual fights at the edges. My strong prediction: firms are going to be badly surprised by how much of this they kept without ever deciding to.
How I’d know I’m wrong. The question getting settled early and cleanly by a rule amendment instead of litigation.
Prediction 11 — The signature doesn’t move
Through 2035 and beyond, every document filed with a court is signed by a licensed human who’s personally answerable for it.
Why. This is the wrapper. The signature isn’t a technical requirement that a good enough system could satisfy. It’s how the legal system figures out who’s responsible. Courts need somebody to sanction. Clients need somebody to sue. Insurers need somebody to underwrite. Bars need somebody to discipline. A system with no answerable person breaks all four at once, and none of them will take that trade for any level of accuracy.
I’d bet against full autonomy in anything with a signature block more confidently than I’d bet on anything above.
How I’d know I’m wrong. Any jurisdiction accepting a filing that can’t be traced to a licensed human. I don’t expect to see it.
Prediction 12 — And the one that isn’t a prediction
None of the above decides this one, and it’s the only one that matters.
Falling production cost doesn’t automatically mean falling prices. It means falling cost. Whether that reaches clients depends entirely on whether competition, regulation, or professional conscience pushes it through. If nothing does, the efficiency turns into margin — same work, same bill, better year.
That’s not cynicism. It’s the default. Efficiency gains go to whoever has the pricing power, and in most legal markets that isn’t the client. A firm that cuts its production cost in half and holds its rates has done nothing for access to justice. It’s just gotten richer, which is fine, and isn’t what any of this was for.
So the real open question about the next decade isn’t technical, and it isn’t really a prediction. It’s a choice the profession makes one firm at a time: whether the gain shows up as a smaller bill and a case you couldn’t afford to take before, or as a bigger distribution.
Everything in this section — the boards, the pipelines, the verification gates, the connectors — is only worth writing down if the answer’s the first one. That’s the whole reason it’s published free, and the only thing this firm asks in return.
Keep score
I tried to write these so you can check them. If you’re reading this in 2035, the ones I expect to have blown are Prediction 2 (I suspect I’m early — defending the process is stickier than the capability) and Prediction 5 (predicting regulators is a mug’s game; the states may just stay split for good).
The one I expect to nail is Prediction 3, and nobody’s watching it.
Sources
- ABA Standing Committee on Ethics and Professional Responsibility, Formal Opinion 512 (July 29, 2024)
- Legal Services Corporation, The Justice Gap (2022)
- Stanford Law School, Regulatory Innovation at the Crossroads: Five Years of Data on Entity-Regulation Reform in Arizona and Utah (June 2025)
- Minnesota Judicial Branch, Legal Paraprofessional Program
- Minn. R. Civ. P. 11.02
- Tremblay v. OpenAI, Inc., No. 23-cv-03223-AMO, 2024 WL 3748003 (N.D. Cal. Aug. 8, 2024)
- Warner v. Gilbarco, Inc., No. 2:24-cv-12333, ECF No. 94 (E.D. Mich. Feb. 10, 2026)
- United States v. Heppner, No. 25 Cr. 503 (JSR), ECF No. 27 (S.D.N.Y. Feb. 17, 2026)
- Morgan v. V2X, Inc., No. 1:25-cv-01991-SKC-MDB, ECF No. 65 (D. Colo. Mar. 30, 2026)
Speculative commentary on the future of the profession. Not legal advice, not ethics advice, and emphatically not investment or career advice. Predictions are the author’s opinion and are offered to be tested, not relied upon.