The largest plaintiff law firm in America is spending $100 million a year on AI โ and the reason has nothing to do with technology and everything to do with how it gets paid.
As reported by Artificial Lawyer, Morgan & Morgan โ the largest plaintiff law firm in America โ has committed at least $1 billion to legal technology and artificial intelligence over the next ten years, roughly $100 million a year, equal to approximately 4% of annual income. This is not a standing start: the firm had already deployed $300 million over the previous five years, making the new pledge a roughly 1.7-fold step-up in annual run-rate against a revenue base of $2.4 billion as of June 2026.
Alongside the commitment, the firm unveiled MX2, a platform described as orchestrating the entire case lifecycle. It is built around four core modules: MX2 Agent, covering case insights, trial preparation, and agentic workflows inside Litify; First Draft, for case-aware document generation with Microsoft Office integration; Medical, handling record extraction, chronology building, and alerts; and Search, providing semantic search across unstructured case data. The system has fulfilled hundreds of thousands of medical record requests and retrieved a substantial volume of police reports.
Other trade coverage โ not the primary Artificial Lawyer write-up โ reports that MX2 has close to 5,000 monthly active users, and that the firm intends to license the platform to other law firms by invitation from late 2027. Both figures should be treated as reported rather than formally confirmed; no licensing terms have been disclosed.
MX2 โ Confirmed Milestones
2021โ2025
$300 million deployed in legal technology and AI โ the foundation on which MX2 is built.
June 2026
$2.4 billion in revenues reported. Reuters, per separate coverage, describes the firm as exploring external investment and eyeing a long-term listing โ characterised by John Morgan as purely exploratory; no transaction announced, no timeline exists.
September 2026
$1 billion, 10-year AI commitment announced. MX2 platform unveiled with four modules. Hundreds of thousands of medical record requests fulfilled.
The key insight: A contingency-fee firm eliminates an hour and keeps the money. An hourly-billing firm eliminates an hour and loses the invoice. That arithmetic โ not engineering talent, not data strategy, not ambition โ is what explains who moves first on automation in law, and almost every other professional-services industry.
A firm that bills by the hour subtracts efficiency from its own top line. A firm paid a share of the recovery adds it to margin. The same project, opposite signs — which is why the biggest number in legal AI came from this side of the profession.
The Structural Read
Co-founder John Morgan did not waste words explaining the commitment:
John Morgan, Co-Founder โ Morgan & Morgan
“Firms which bill by the hour…are the practices AI will replace almost entirely.”
This framing is often read as competitive bravado, but the more useful reading is structural. Hourly-billing firms are not slow because of culture or caution. They face a specific arithmetic problem: every hour of work automated is an hour removed from the top line. A partner approving an automation project inside a billable-hour partnership is, in accounting terms, approving a reduction in billable inventory. That is a genuinely difficult thing to champion when profit is distributed annually and the subtraction is immediate. This does not imply that such firms are poorly run or incapable โ it means they are solving a different equation, and it would be strange if they reached the same answer.
Morgan & Morgan operates on contingency. It collects a percentage of the recovery regardless of the hours consumed. Every hour eliminated by MX2 falls straight through to margin. The identical technology that is subtractive on one income statement is purely accretive on the other. The commitment is large because the incentive is aligned, not because the firm is uniquely far-sighted.
Portable Principle โ Revenue Model Governs Adoption
Time-sellers vs. Outcome-sellers
Where the unit of sale is time, automation cannibalises the seller โ adoption is slow, defensive, and confined to work the client cannot see. Where the unit of sale is an outcome, automation expands margin โ adoption is fast and enthusiastic. Time-sellers include agencies, consultancies, accountancies, and traditional law firms. Outcome-sellers include contingency litigation, SaaS subscriptions, insurance, and asset management. The adoption curve tracks that division more reliably than it tracks model capability or engineering headcount.
The second layer of this story concerns what the billion dollars is actually building. Model access is not the scarce input in vertical AI โ anyone can purchase it at list price. The scarce input is proprietary operational data, and a plaintiff firm operating at Morgan & Morgan’s scale generates exactly that as exhaust. Hundreds of thousands of medical record requests, police reports, court dockets, and client interactions are produced as a by-product of running cases, not assembled deliberately. No legal-technology vendor can replicate an equivalent corpus through licensing, because the corpus only exists where the casework happens. The four MX2 modules are less finished products than instruments for converting case flow into structured data โ and that conversion, compounding over years, is the genuinely defensible component of the investment.
A fourth observation should be labelled clearly as this publication’s interpretation rather than reported fact. Law-firm earnings do not attract software valuation multiples. In most United States jurisdictions, the ownership of a law practice by non-lawyers is restricted โ though the rules vary by jurisdiction and a small number of states have moved differently; nothing here constitutes legal advice. That restriction is why the vehicle appearing in the reporting is the technology and case-management operation rather than the practice itself. Building MX2 inside the firm, training it on the firm’s own case flow, and then separating it would convert law-firm earnings into software earnings. On that reading, the $1 billion technology commitment and the separately reported exploration of a billion-dollar-plus minority stake โ which Reuters described in speculative rather than confirmatory terms in June 2026, and which John Morgan has characterised as purely exploratory โ could be understood as two ends of a single structural transaction. To be unambiguous: no transaction has been announced, no timeline exists, and no offering size or valuation has been confirmed. This is one analytical reading of publicly reported and explicitly exploratory discussions, and nothing here predicts that any sale, license, or listing occurs.
The most telling detail is the reported licensing plan itself. Opening MX2 to rival firms by invitation only makes sense if you believe the compounding value of data scale exceeds whatever competitive advantage you surrender by doing so. A law firm would not reason that way about its case-handling edge. A software company reasons that way by default. That choice โ not the size of the number โ is the signal worth tracking.
Before asking who will automate next in any professional-services sector, ask whose revenue is decoupled from hours. That question produces a more reliable answer than surveying engineering budgets or AI strategies. Contingency litigation moved first in law. The equivalent actors in accounting, consulting, and financial advice are identifiable by the same test โ they are already carrying the incentive, whether or not they have yet acted on it.
2 โ PROPRIETARY OPERATIONAL DATA IS THE DURABLE MOAT
The scarce input in vertical AI is not model access โ it is the data generated as a by-product of doing the work at scale. Firms that have been running high-volume operations for years hold a corpus that outside vendors cannot assemble by licensing. The implication for incumbents in any data-rich vertical: the window to convert operational exhaust into structured AI training data is not indefinitely open, and the compounding advantage accrues to whoever starts the conversion earliest.
3 โ LICENSING TO COMPETITORS IS SOFTWARE LOGIC, NOT LEGAL LOGIC
If the reported late-2027 invitation-only licensing plan proceeds, it will confirm which business Morgan & Morgan believes it is in. A law firm protects its case-handling edge; it does not sell it to rivals. A software company sells to rivals when network effects and data scale outweigh the competitive cost. The licensing decision, if taken, is the clearest available signal that MX2 is being built as a separable asset โ and that the firm’s long-term earnings ambition lies in the software column, not the legal-services one.
Morgan & Morgan is executing a classic Founder-to-Distributor transition: it built proprietary case-flow data as a Founder operating in a vertical, and is now positioning MX2 as a Distribution layer for the broader legal market. The FDE Framework maps exactly where a company sits in the AI stack โ and, critically, which position commands durable margin. Understanding whether an incumbent is a Founder, Distributor, or Enabler determines how to read moves like this one.
91,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
No transaction has been announced. Reporting on external investment, a minority stake sale or an eventual public listing is explicitly exploratory: no timeline exists, no terms have been disclosed, and John Morgan has characterised the discussions as purely exploratory. Nothing here states or implies that any sale, investment or listing has been agreed, planned or scheduled, and no valuation or offering size is asserted. The $1 billion is a commitment of at least that amount spread over ten years, not a sum already spent. The figure of approximately 5,000 monthly active users for MX2, and the plan to licence the platform to other firms by invitation from late 2027, come from trade coverage other than the primary write-up and are reported rather than established. No claim is made here about MX2’s performance or results beyond what has been reported. The reading that the technology build and the reported stake discussions form two ends of a single transaction is this publication’s interpretation of public reporting, not a described plan. Restrictions on the ownership of law practices by non-lawyers are described in general terms only. Those rules vary by jurisdiction, no individual jurisdiction’s position is stated here, and nothing in this article is legal advice or a statement of what any rule requires in any particular case. The analysis of billing models is structural and concerns incentives rather than competence. Nothing here suggests that firms billing by the hour are poorly run, slow to adapt or inferior, and no firm is named or characterised as a laggard. Morgan & Morgan is a private partnership; Microsoft is publicly listed; Litify, Supio, EvenUp, Eve and NewMods are private companies. This is businessanalysis. It is not investment advice and not legal advice, no view is expressed on any security, and no recommendation is made.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
