The provocation was calculated, at the end of last week, on the stage of a London conference, Max Junestrand, founder and CEO of Legora, issued a sentence intended to become a landmark: “Legal AI is dead. » Behind the laconic formula, the observation is that the first generation of legal artificial intelligence has lived, now in place of “Legora aOS”, an operating system designed to entrust AI agents with the execution of complete legal tasks, with human intervention reduced to what is strictly necessary.
The end of the co-pilot, the advent of the agent
In a particularly active legaltech market in recent months, Max Junestrand is directly targeting the wave of startups born in the wake of ChatGPT. For two years, the market was structured around conversational assistants capable of summarizing contracts, speeding up legal research or drafting documents using prompts. If these tools improved the individual productivity of the lawyer, they in no way called into question the deep organization of the firms.
It is on this point that Legora intends to move AI from the status of an additional tool to that of operational infrastructure. Its system functions as an environment capable of orchestrating end-to-end legal workflows, where agents analyze contracts, compare clauses, conduct regulatory research, prepare responses and handle administrative tasks on an ongoing basis. The goal is to automate entire execution chains, not just to help a lawyer work faster.
The company describes its OS as “a unified and connected system which facilitates the circulation of information, exchanges and the execution of legal work”, a unified layer intended to centralize all of a firm’s document flows, exchanges and operations.
When the agent works while the firm sleeps
The example put forward by Legora illustrates the extent of this ambition. When a counterparty submits a marked-up version of a contract in the middle of the night, the agent reviews changes, identifies risks, develops counterproposals, and transmits a pre-written response before teams even arrive the next morning.
A model that does not make the lawyer disappear but profoundly transforms his role. He becomes a supervisor, strategic arbiter and quality controller of a system which itself produces a substantial part of the legal work.
“Agent Legora is the most powerful AI we have developed”says Max Junestrand. “Until now, the ceiling of what could be done was limited by human capacity. The Legora aOS changes that. This is the operating system that enables legal teams to operate with machine intelligence at a scale, speed and quality that was simply impossible before. »
The pyramid model in the crosshairs
This transformation touches the economic heart of large international firms. For decades, law firms have been based on a pyramid logic, partners, senior associates, juniors and interns, in which a considerable part of revenue comes from tasks that consume a lot of human time: documentary review, compliance, preparation of regulatory responses, contractual negotiations.
However, it is precisely these operational layers that AI agents target. The issue therefore goes beyond the sole question of productivity: it involves the economic model itself. If autonomous systems absorb an increasing share of the work historically entrusted to junior employees, several questions immediately become critical. How to continue billing hours on automated tasks? How to preserve margins? And above all, how can we train future associates if the first years of apprenticeship, those where we learn the profession through practice, gradually fade away?
This tension explains why the entire sector now insists, with almost unanimous caution, on the notion of human supervision. At Legora as at Harvey AI, the official discourse remains that of an AI supervised by legal professionals. But behind this regulatory restraint, the market is moving towards a methodical industrialization of legal work.
A dizzying valuation race
The speed with which this battle is escalating is astonishing. Founded in 2023 in Stockholm, Legora is now one of the most valuable companies in the European AI sector. In March 2026, the company closed a $550 million Series D round led by Accel, bringing its valuation to $5.55 billion. A few weeks later, Nvidia, via NVentures and Atlassian injected an additional 50 million into an extension of the round, propelling the valuation to around $5.6 billion.
Legora’s financial trajectory is dazzling: the startup was worth around $675 million in spring 2025, reaching $1.8 billion a few months later, before crossing the $5.5 billion mark at the start of 2026.
Harvey, however, remains one step ahead, the American company, supported by Sequoia, a16z, Coatue, Elad Gil and Kleiner Perkins, has reached a valuation of $11 billion after a new round of $200 million in early 2026. Harvey claims more than 100,000 user lawyers spread across some 1,500 firms and legal departments, with several hundred thousand tasks performed daily by its agents.
A market in the process of fragmentation
The confrontation is no longer reduced, however, to a transatlantic duel between two behemoths. The legal AI market is starting to stratify itself into several categories of platforms. Harvey and Legora aim to become the operating systems of large international firms. Other players position themselves on specific functional layers: Luminance on contractual analysis, Robin AI on negotiation and document management, Spellbook on integrated contractual assistance, Wordsmith AI on agentic legal workflows.
In continental Europe, historical players nevertheless have a major advantage: mastery of local legal corpus and national customs. In France, Doctrine has built a dominant position in augmented legal research and the exploitation of court decisions, thanks to a massive documentary base and a solid presence among firms and legal departments. Predictice, now integrated into Septeo, has established itself in case law analysis and litigation decision support. DiliTrust, strengthened by the acquisition of Hyperlex, targets legal departments with an approach focused on contract managementdocument governance and compliance workflows. Legalstart occupies a distinct segment, that of automated legal formalities for SMEs and entrepreneurs. Players like Gino LegalTech or Case Law Analytics are exploring document automation, predictive analysis or litigation simulation.
These companies do not compete in exactly the same game as Harvey or Legora, but they hold eminently strategic assets with proprietary documentary bases, established relationships with legal professionals and in-depth knowledge of European regulatory environments. The central question is now asked: will these platforms manage to evolve into complete agentic systems, or will they be absorbed into more global AI infrastructures?
Legal data, the sinews of war
Legora specifically attempts to prevent this fragmentation by controlling not only the AI interface, but also the operational memory of the cabinets. The real strategic asset of the sector is no longer the algorithmic model but the legal data itself.
The system is designed to leverage firms’ internal knowledge, precedent libraries, banks clausecustomer histories, negotiation positions, internal methodologies, playbooks sectoral. Each agent-generated response should reflect how the organization has historically handled comparable situations. Legora emphasizes technical capabilities specifically trained for law: legal research contextualized according to jurisdiction, structured documentary analysis on thousands of files, semantic understanding adjusted to legal uses.
The company thus seeks to erect a layer of operational legal memory that is difficult to reproduce and, thereby, to lock in the adherence of its customers.
The rise of “legal engineers”
This development explains the growing importance of an unprecedented profile in the legal ecosystem: the legal engineerat the crossroads of legal expertise, automation and technological integration. Legora claims to now employ as many legal engineers than software engineers.
Their role goes far beyond that of traditional technical support. These teams are integrated into client projects to configure workflows, adapt knowledge bases and translate the firms’ working methods into the operational functioning of the aOS. The profile is similar to that of forward deployed engineers popularized by Palantir Technologies: intermediaries capable of adapting AI systems to the real operational constraints of organizations.
The main challenge is no longer solely algorithmic but organizational, namely structuring data, standardizing workflows, connecting tools, managing compliance, transforming the operational habits of firms that are often resistant to change.
A change that goes beyond legaltech
After software development, customer support, cybersecurity and finance, law is in turn becoming a testing ground for agentic systems. Regulated professions, historically protected by technical complexity, scarcity of skills and human cost, are entering a phase of cognitive automation, the extent of which no one can yet measure.
In this context, the best technologically structured firms could benefit from a considerable advantage. Those who know how to exploit their internal knowledge, standardize their workflows and quickly integrate these new infrastructures will be able to absorb more work with smaller teams and significantly reduced operational costs.
“Agentic Law is a fundamentally different model of how legal work is done and what firms can accomplish”estimates Max Junestrand. “The firms that move first will define how law is practiced for the next decade. »
What is at stake now goes beyond the adoption of an additional tool. It is the transformation of law firms into organizations capable of delegating an increasing part of their operations to autonomous software infrastructures, powered by their own data, internal methods and accumulated knowledge. Law no longer escapes the law of automation. It remains to be seen who, lawyers or machines, will write the next rules.
