Facts are friends
About this series
This essay is part of a public debate Lev Loukhton and I are having. We are calling it the #TrojanHorseDebate.
Lev and I came to this conversation from different starting points. We have discovered that there are many things on which we do not agree. But we are happy to disagree well and share genuine interests in following the argument wherever it leads, without commercial positions to defend or conclusions to protect. We are both in the midst of the Legal AI world as it is being built, and we think that the questions being raised deserve more rigorous public examination than they are currently getting.
This debate began when I published my five part Substack series on Legal Tech valuations being built on sand, leading to the assertion that the Legal Tech solutions were Trojan Horses to law firms (links below). Lev disagreed, and did not hold back in saying so. Rather than a public gladiatorial style debate we engaged with one another meaningfully. And we found that whilst disagreeing, we respected one another. It is a good dynamic.
Then Lev (with my agreement) published an argument on LinkedIn mentioning me. He anchored his post in a story about Michael Gerstenzang of Cleary Gottlieb presenting his leadership team with a challenge nine years ago: imagine the business that puts Cleary Gottlieb out of business, and build it. Lev’s argument is that this spirit of adaptation is the legal profession’s answer to AI disruption. His post is here. My LinkedIn response is linked to in my comment on his post. This Substack essay is the analytical case behind my response.
The next move is Lev’s. I am intrigued to see what he says and how this develops. I hope you have a chance to form your own view before reading where we go.
You are welcome to join the debate in the comments. We mean it - Lev and I are interested in what you have to say.
Quentin Solt, March 2026
The challenge
Lev opened this public debate with his post based on a story told by Michael Gerstenzang, Cleary Gottlieb’s Senior Partner, who presented his leadership team nine years ago with a challenge:
“imagine the business that puts Cleary Gottlieb out of business, and build it”.
The result was ClearyX - an AI and data-driven delivery unit that has since worked alongside Cleary on over 150 matters, staffed primarily by legally trained analysts, data scientists, project managers and software developers rather than billing associates.
Lev’s argument is that this is adaptation. That the profession has always adapted - to the post-GFC squeeze, to client sophistication, to the red flag report replacing the 500-page due diligence bible, and that it will adapt again. His predication and conclusion:
“little cause for structural alarm”.
I agree that adaptation is happening. What I want to examine is where the economics of that adaptation end up - and whether the analogy for what Cleary and others are doing is adaptation, or something that looks rather more like Kodak.
The Kodak problem
In 1975, a Kodak engineer named Steve Sasson built the world’s first digital camera.
Kodak’s management saw the prototype, understood exactly what it would do to the film business, and made a calculated decision: develop it carefully, contain it, and use it to protect the core franchise rather than replace it. They were not ignorant of the threat. They built the technology. They held the patents. Just like Michael Gerstenzang was prescient in seeing the AI future of law.
Unlike Kodak which ceded the digital market to Sony, Canon and (eventually) the smartphone, while protecting a business the market rendered obsolete anyway; Cleary brought the fight home by developing its own challenger - ClearyX. The hope must be to avoid what happened to Kodak - it filed for bankruptcy in 2012. The business that could put it out of business had been sitting in their own R&D lab for nearly four decades. Doubtless ClearyX is designed to avoid such a disaster for Cleary Gottleib.
By building ClearyX the firm which sees the model that could undercut its core economics, builds it internally, and reasons that controlling the disruption is better than conceding it to someone else. That reasoning is not wrong. The question is whether control is the same as capture.
The newspapers made the same calculation in the late 1990s. They built digital editions, often giving content away free to capture audience. They understood that online classifieds and digital advertising were coming. Several modelled the cannibalisation risk explicitly. They built it anyway. Digital revenues never replaced print revenues. The economics of journalism collapsed. They built the business that put them out of business, and they did it with full awareness of what they were doing.
The recording artists’ labels licensed their catalogues to iTunes in the early 2000s as a controlled digital distribution channel. They trained their customers to expect à la carte purchasing at 99 cents a track, destroyed the album as an economic unit and handed distribution leverage to Apple. Spotify completed the transition. The labels now receive roughly 58 pence in every £ of streaming revenue compared to what physical media had generated. The labels licensed themselves into a structurally weaker position. I am on the board of OpenStage, which is carefully allowing artists to regain control of their own fans - but the old model is lost forever (and that - unless you relied on the old model to make money - is not necessarily bad, but it is not what the incumbents expected or would have planned for).
The pattern across each of these is the same. The incumbent sees the disruptive model clearly. It builds a version of it to retain control. It cannot escape the structural economics of the new model because its existing cost base, margin expectations and organisational incentives are calibrated for the old one. That is the problem I see with Lev’s optimistic assertion that the profession adapts: it does - but that may not be enough.
Who owns the factory?
It is clear that Cleary holds the title deeds to the ClearyX, the “new factory”. My question is whether ownership of a lower-margin delivery vehicle compensates for the structural compression it creates in the higher-margin business that funds the partnership.
ClearyX delivers M&A due diligence at 30-60% lower cost than traditional associate review. That efficiency is real and the client benefit is genuine. Cleary faced a binary choice: lose the work to cheaper competitors or build ClearyX to retain it.
But rational choices at firm level do not prevent structural damage at profession level.
Every dollar retained through ClearyX at a 30-60% discount is multiple dollars that previously fed through to partner profits. The model that funds the BigLaw pyramid - junior associates billing at $350-550 an hour, generating the spread that makes equity partner profits possible - gets thinner each time the factory runs.
A partnership with equity partners earning $3-5 million a year cannot run a technology platform business. The two models require fundamentally different capital structures, talent models and pricing philosophies. The margins on process-optimised, technology-enabled delivery are structurally different from the margins on judgment-intensive, partner-led practice.
Let me explain by reference to another industry, airlines. Every major European flag carrier that launched a low-cost subsidiary between 1998 and 2004 either sold it, shut it down or watched it fail to compete on cost with Ryanair and easyJet.
British Airways sold Go for £100 million - then watched easyJet buy it for £374 million a year later.
KLM unloaded Buzz to Ryanair for a net cash cost under €5 million.
SAS killed Snowflake after just 19 months.
Lufthansa's Germanwings/Eurowings survived in name but has never achieved genuine low-cost economics, operating at roughly double Ryanair's unit cost.
The budget carriers these subsidiaries were designed to neutralize now collectively form the largest market segment in European aviation at 35.2% of all flights, having overtaken legacy mainline carriers. Ryanair alone carried 206.5 million passengers in 2025 more than the Lufthansa Group's entire operation. No European legacy carrier has ever built a low-cost subsidiary from scratch that genuinely competes with standalone budget airlines on unit cost. Explain to me why law is different.
This is not a criticism of ClearyX. I am not deeply familiar with it, but from what I see it is a serious, well-executed piece of institutional engineering. But I do not doubt that the same would have been said of the flag carriers’ creations in the low-cost market place. The criticism is not managerial, it is structural. Starting from a position of dominance, with market expectations and historic architecture and overheads, brings legacy drag.
The platform moves faster
Now step one level further. Harvey AI confirmed an $8 billion valuation in December 2025, growing revenues roughly four times in a single year. It sells to over 50 Am Law 100 firms. It simultaneously sells directly to those firms’ corporate clients. Its own research shows that in-house teams using Harvey are actively reducing their external counsel spend - in some cases by 5-10%. The Adecco Group reported reducing reliance on outside counsel. Syngenta’s legal team described running issues through Harvey before deciding whether to instruct external counsel at all.
Harvey’s platform serves both sides of the relationship simultaneously. The law firms using Harvey to deliver better work are also, through that same adoption, training their clients to need them less. This is not a conspiracy. It is the natural logic of a two-sided platform capturing value from both sides of a market. It is a part of the basis of my Trojan Horse argument that Lev so riled against.
Gerstenzang, whose candour throughout this debate has been striking, put it succinctly:
“Generative AI is going to affect every practice and every person. The only question is the time frame, not the direction”.
He is right. And that framing contains something important that I argue Lev’s adaptation argument does not fully address. Adaptation manages the direction of change. It does not resolve the question of where the value ends up.
The scarcity problem
There is a deeper issue beneath the platform economics.
The legal profession’s commercial value has always rested on scarcity - of expertise, of access, of trained judgment. A partner at a leading firm commands premium fees because of what they know and because the ability to do what they do is genuinely scarce. That scarcity is not accidental. It is the product of a training pipeline that takes decades to produce a senior practitioner, of regulatory barriers that limit who can practise, of the sheer complexity of the work at the top of the market and of the limited availability of the safety net of insurance.
What happens to scarcity when the tools that encode legal expertise become widely and cheaply available? What happens to the moat of insurance when the insurers trust the competition as much as they trust the incumbent. What happens to the incumbent when the regulatory protectionism becomes a wall to climb before you can access your client?
The firms deploying Harvey, Legora, Clio, CoCounsel and their myriad alternatives and their respective successors are, in aggregate, teaching the market that sophisticated legal analysis can be produced faster and cheaper than it previously could. That is genuinely good for access to justice. It is also, over time, corrosive to scarcity - one of the tenets creating the premium that today’s sophisticated legal market commands.
They will succeed in commoditising their speciality. That very success removes the scarcity which has previously commanded the value.
I am aware of the obvious response to this argument: what is the alternative? If refusing to participate is not viable - and with clients actively demanding AI adoption, it is not - then the diagnosis changes nothing. That response deserves a direct answer. I am not arguing that law firms should have done something different. I am arguing that the profession needs to see clearly what it is walking into. The Prisoner's Dilemma is real: no individual firm, acting alone, can preserve the scarcity that commands premium fees. Each firm's rational individual decision accelerates the collective outcome. Understanding that dynamic does not change the individual calculus. But it should change how the profession plans for what comes next - and that planning is largely absent from the current debate, which remains dominated by the reassuring language of adaptation.
What remains
I am not arguing that lawyers disappear (at least not immediately). I am arguing that the economics of legal practice are being structurally reorganised, and that the firms best placed to capture value from that reorganisation are the technology platform operators, not the partnerships.
What survives for now is the commercial advisory layer. The senior practitioner who understands a client’s business, who can navigate a complex regulatory environment, who can read a negotiation and advise on strategy rather than process.
I believe that that judgment is automatable, but not in the near-term horizon. It is also, critically, not what most of the billable hours in a large firm currently represent. It is, however, the value that justifies those hours. A sort of bridge that clients need to cross to access the gold. It is, also, I argue as much a consultancy role as a legal role.
The legal profession is not yet building its extinction. That may follow. But it is building a future in which a smaller number of very well-rewarded practitioners offer genuinely scarce commercial judgment, supported by technology platforms that capture the process economics. The question is whether that future looks like the legal profession as we know it today - or whether it looks like something quite different, with the economics flowing elsewhere.
Kodak phoenixed , eventually. As a shadow of its former self. Some of today’s legal brands may similarly survive.
Quentin Solt is a UK solicitor and legal consultant. My forthcoming book, AI, Abundance and the Future of Work, explores how AI reshapes society, work, politics and human connection. I writes the #FactsAreFriends series on Substack.
This essay is part of a public dialogue with Lev Loukhton at #TrojanHorseDebate.
References are included below, but first the posts that launched the debate:
__________________
References
ClearyX - launch, structure and economics
Cleary Gottlieb official ClearyX launch announcement: https://www.clearygottlieb.com/news-and-insights/news-listing/cleary-gottlieb-launches-clearyx
Artificial Lawyer - Carla Swansburg “whole new factory” quote, ClearyX staffing model and founding vision (February 2022): https://www.artificiallawyer.com/2022/02/28/cleary-x-will-be-an-ai-driven-transactions-factory/
ClearyX transactional diligence - 30-60% cost reduction claim: https://clearyx.legal/cleary-gottlieb-launches-clearyx-a-platform-for-highly-efficient-ai-and-data-driven-legal-services/
Artificial Lawyer - Daniel Himmel interview on ClearyX evolution and AI deployment (May 2024): https://www.artificiallawyer.com/2024/05/09/daniel-himmel-clearyx-on-genai-the-groups-evolution/
Computerworld - Cleary Gottlieb GenAI deployment in practice including ClearyX technology stack: https://www.computerworld.com/article/3564813/heres-how-cleary-gottlieb-law-firm-uses-genai-for-pre-trial-discovery-and-more.html
Michael Gerstenzang
Sixth Street “It’s Not Magic” podcast - Gerstenzang on AI reducing lawyer headcount and the direction of change: https://sixthstreet.com/podcasts/michael-gerstenzang-managing-partner-of-cleary-gottlieb-steen-hamilton/
The Deal - Drinks With The Deal podcast - Gerstenzang on AI affecting every practice and every person: https://www.thedeal.com/podcasts/drinks-with-the-deal-clearys-michael-gerstenzang-talks-ais-effect-on-legal-practice/
Original Jurisdiction - David Lat interview - Gerstenzang “windmills not walls” and AI transformation philosophy:
Bloomberg Law - Gerstenzang on the billable hour becoming “completely antiquated”: https://news.bloomberglaw.com/business-and-practice/cleary-makes-generative-ai-play-with-springbok-legal-tech-buy
Harvey AI - platform economics and two-vector strategy
TechCrunch - Harvey AI confirms $8 billion valuation, December 2025: https://techcrunch.com/2025/12/04/legal-ai-startup-harvey-confirms-8b-valuation/
Sacra - Harvey revenue, funding and growth data: https://sacra.com/c/harvey/
Legal IT Insider - RSGI/Harvey adoption report: in-house teams reducing external counsel spend: https://legaltechnology.com/2025/12/02/the-impact-of-legal-ai-a-deeper-dive-into-the-rsgi-harvey-adoption-report/
Harvey AI - Syngenta case study: running issues through Harvey before instructing external counsel: https://www.harvey.ai/customers/syngenta
Harvey AI - Adecco Group case study: reducing reliance on outside counsel: https://www.harvey.ai/customers/the-adecco-group
Harvey AI - “A new era of collaboration for legal and professional services”: https://www.harvey.ai/blog/a-new-era-of-collaboration-for-legal-and-professional-services
Harvey AI - “What AI adoption reveals about the future of legal work”: https://www.harvey.ai/blog/what-ai-adoption-reveals-about-the-future-of-legal-work
The captive ALSP structural argument
Bruce MacEwen, Adam Smith, Esq. - “Captive ALSPs are a dead end” - the Song and Ted airline analogy and why parent firms cannot sustain fundamentally different business models: https://adamsmithesq.com/2021/10/captive-alsps-are-a-dead-end-discuss/
Thomson Reuters and Georgetown Law - 2025 report declaring 2024 “the beginning of the end for the traditional law firm business model”: https://www.lawnext.com/2025/01/report-from-thomson-reuters-and-georgetown-law-says-2024-was-the-beginning-of-the-end-for-the-traditional-law-firm-business-model.html
Historical analogies
Kodak and the digital camera - Steve Sasson’s 1975 invention and Kodak’s managed decline: https://tms-outsource.com/blog/posts/what-happened-to-kodak/
The newspaper industry and digital cannibalisation - background: https://www.niemanlab.org/2023/12/the-newspaper-industrys-slow-decline/
Am Law 100 financial data
Am Law 100 2024 financials - revenue, PPP and headcount:
European airline analogies
British Airways Go Fly - history, launch and sale to easyJet: https://en.wikipedia.org/wiki/Go_Fly
Simple Flying - what happened to BA’s Go Fly subsidiary: https://simpleflying.com/british-airways-go-fly-fate/
Airways Magazine - the full history of Go Fly: https://www.airwaysmag.com/legacy-posts/the-history-of-go-fly
Simple Flying - when easyJet gatecrashed BA’s Go launch: https://simpleflying.com/throwback-easyjet-gate-crashes-british-airways-go-subsidiary-launch/
KLM Buzz sold to Ryanair: https://www.key.aero/forum/commercial-aviation/6754-klm-sells-buzz-to-ryanair
SAS Snowflake - failed after 19 months: https://en.wikipedia.org/wiki/Snowflake_(airline)
Flight Global - Snowflake profit problems: https://www.flightglobal.com/snowflake-faced-with-meltdown-over-profit/53904.article
Lufthansa/Germanwings/Eurowings evolution - CAPA Centre for Aviation: https://centreforaviation.com/analysis/reports/lufthansa-to-germanwings-to-eurowings-long-haul-and-lower-cost-as-lufthansa-seeks-solutions-255083
Eurowings unit cost versus Ryanair - CAPA analysis: https://centreforaviation.com/analysis/reports/eurowings-develops-innovative-partnership-approach-but-cask-reduction-may-not-go-far-enough-287300
Ryanair 206.5 million passengers in 2025: https://www.irishexaminer.com/business/companies/arid-41769905.html
Ryanair now UK’s biggest airline with double BA’s traffic - CAPA: https://www.routesonline.com/suppliers/10554/capa-centre-for-aviation/news/299665097/ryanair-uks-biggest-passenger-airline-with-double-bas-traffic/
Low-cost carriers now largest segment in European aviation at 35.2% of flights - EUROCONTROL: https://www.eurocontrol.int/publication/eurocontrol-data-snapshot-34-rise-low-cost-carriers-europe
Academic analysis: incompatibilities of low-cost and network carrier business models within the same airline group (Graf, 2005): https://www.sciencedirect.com/science/article/abs/pii/S0969699705000578