In 1865, the British Parliament passed the Locomotive Act. Better known as the “Red Flag Act,” this law required self-propelled vehicles to travel at no more than two miles per hour in towns and four in the countryside. They were required to be crewed by three people, one of whom had to walk ahead and wave a red flag to warn approaching horse-drawn traffic. The law’s purpose, it was stated, was safety. Its real purpose was to protect the horse-drawn economy from the machine that would replace it.
The Act held for three decades. It did not save a single horse-drawn business. By the time the automobile prevailed over the horses, however, Britain’s early advantages in automotive engineering had been surrendered to France, Germany and the USA.
Today’s legal profession is living through its own Red Flag moment. The “horses” are the established, people-leveraged business model that has served law firms so well for so long. The “automobiles” are emergent, AI-native legal advisory businesses. These represent a fundamentally different way of delivering legal services, built from scratch around AI rather than being slotted into teams of human fee-earners. Their economic drivers differ radically.
What the AI-native firm looks like – and why clients will prefer it
The defining feature of an AI-native law firm is not that it uses AI. Every firm already does that, whether they realise it or not. It is that the firm is designed, from inception, around a fundamentally different ratio of people to output. This results in very different economics. Consider what is already happening in adjacent industries. Cursor, the AI coding tool, reached $100 million in annual recurring revenue with roughly 20 people. So roughly $5 million per employee. Midjourney, an AI art generator, generated $200 million with 40 people. Lovable, an AI app builder, hit $200 million in eight months with 15 people. Very different businesses to law firms … yes, of course I know. But any traditional services company considers $300,000-$500,000 revenue per employee (not just per lawyer) excellent. AI-native firms run at five to seven times that figure. The gap is structural, not marginal. It represents a phase change in business operating model.
In these disruptive AI native firms, small numbers of skilled humans orchestrate fleets of AI agents, each of which can perform tasks that would previously have required multiple people. The humans set direction, evaluate quality and make the calls that require genuine expertise. The agents execute, coordinate and scale. The leverage has shifted from execution to judgement. Now imagine that architecture applied to legal services.
To reiterate: AI-native law firms do not employ ranks of associates to review documents, draft standard agreements or research precedent. Instead, AI agents do that faster, at a fraction of the cost, and more consistently and accurately than junior lawyers can. Their senior lawyers do what clients actually value most: exercising judgement on complex, novel problems; counselling on risk; and helping clients navigate situations where the answer is not in the textbook. Everything else – the vast middle ground of competent legal work that currently sustains the leveraged law firm model – is delivered by agents under expert human supervision.
But what, in this model, becomes of junior lawyers? Their role changes profoundly. Rather than spending years on the repetitive tasks that have traditionally served as an apprenticeship, they learn to direct and supervise AI agents from the outset. Much like a junior architect uses CAD from the outset and never touches a drawing board. Required skill sets shift from execution to orchestration: specifying tasks with precision, evaluating the quality of AI output and knowing when, and how, the agent’s work requires human intervention. This kind of work is furthermore, very arguably, a far more attractive prospect for a junior lawyer.
Tools like Anthropic’s Cowork already allow non-technical professionals to dispatch complex tasks by describing the outcome rather than prescribing the steps. A junior lawyer trained to work this way develops judgement faster, not slower, because s/he is exposed to a far greater volume and variety of work than the traditional model permits. Of legitimate concern though is that the “grunt work” was also where institutional knowledge was absorbed through thousands of small exposures. AI-native firms must solve this deliberately, through structured mentoring, closer supervision of higher-value work and early exposure to client-facing situations. Firms that treat this as an afterthought will struggle to develop the senior lawyers they will need in a decade (at most). Those that design for it will build a formidable talent pipeline.
From a client’s perspective, the difference offered by AI native firms is decisive on every dimension that matters to them. Work that traditionally takes a team of associates a week takes an agent minutes (so, speed). The pricing of services reflects the actual cost of delivery, not the cost of an army of fee-earners (so, cost). AI agents do not have bad days, do not cut corners under time pressure, and apply the same standard to the thousandth document as to the first (so, quality). A firm structured this way can afford to serve clients and matters that a traditional leveraged model would consider too small to be profitable (so, access).
The fee pressure this creates is already visible in adjacent professions. When KPMG pressured Grant Thornton to cut audit fees on the basis that AI had reduced the real cost of audit work, Grant Thornton’s audit fees for that client fell 14% in a single year – from $416,000 to $357,000. Importantly, KPMG’s audit was not then automated by AI. The firm merely used the existence of AI as a negotiating lever. The implicit message of “we both know this work costs less now, so your old prices are no longer justified” is one that many general counsel are already delivering to their external counsel. Law firms, predictably, are resisting this message. But they have a simple, binary choice: death by a thousand cuts, or leapfrog to a new and radically different AI native model (perhaps one resembling that of Cursor/Midjourney/Lovable and many others.)
The pattern echoes in other fields. Newspapers had content people wanted. The internet destroyed neither that content nor demand for it. What was destroyed was the assumption that consumers would pay for a bundled product and that advertisers had no alternative. The content survived. The business model did not. McKinsey is already acting on this logic, targeting parity between AI agents and human consultants across the firm by the end of 2026. Dario Amodei, CEO of Anthropic, has placed the odds of a billion-dollar solo-founded company emerging by the end of 2026 at 70–80%. The relationship between headcount/scale and output is breaking. Law firms that grasp this reality and have the strength of leadership to act will redesign around it. Those that do not will find their pricing power eroded, likely quickly.
Organisational ambidexterity: Exploit and explore
This is all easier said than done. The threat is clear but the response is less so. Established law firms cannot simply shut down their current operations and reopen as AI-native businesses. They have partners, staff, clients and obligations. Transitions must be carefully managed.
This is where what management scholars call “organisational ambidexterity” becomes critical. This concept, developed by Charles O’Reilly of Stanford and Michael Tushman at Harvard, addresses a key issue that Clayton Christensen’s theory of disruptive innovation identified but left partly unsolved: How can incumbents respond to tomorrow’s disruption without destroying what sustains them today? Christensen showed that successful firms fail not because they are badly managed, but quite the opposite. They are optimised so thoroughly for their current business that they cannot pivot to a fundamentally different one, even if they want to. Ironically, the more successful a business has been under an existing paradigm, the more difficult it is to shift to a new one. Powerful forces within the firm that benefit from the existing order will also be quick to warn against “frightening the horses” and will tend to try to shut down initiatives that threaten that order.
O’Reilly and Tushman proposed a structural answer. The firm must simultaneously “exploit” its existing business and “explore” the new one, housing them in separate units with different cultures, processes and metrics, but linked at the top by a senior leadership team committed to both. Constructive disruption in the long term interests of the firm’s survival must be their strategic intent.
Low power innovation committees and teams focused on implementing incremental innovations are good for cheerleading but not much more. The people-leveraged model has created decades of extraordinary prosperity. Partner compensation systems are hardwired to billing human effort, measured in time. Inertia is an immensely powerful obstacle. So too that disruptive innovations are in early stages not very good, focused on the fringe, not what clients want. Partners who discount them on that basis (implying that they will remain so) seldom realise that they are reacting precisely as Christensen’s theory predicts. They are right …. until the disruptive innovation improves to the point that they are proved wrong. In today’s world, that can take months or a few years, but one cannot assume that it will take longer.
Every sprint that a law firm innovation team undertakes to add marginal efficiency to its existing model is a sprint not undertaken to build the new one. Bolting AI onto existing workflows (a chatbot here, an AI review tool there) will yield useful incremental gains, but it will not protect a firm from a competitor that has designed, from the ground up, new ways to deliver legal services at five times the revenue per head. Self-disruption is undoubtably the hardest form of innovation, but today’s market demands no less. It requires leaders to cannibalise their own model before someone else does. In law firms, where the partnership structure gives every incumbent partner a vote – and an economic interest in the status quo – the socio-political challenge is as formidable as the strategic one.
The flag or the road
The Red Flag Act bought the British horse-drawn economy thirty years …. but did not save it. Law firm leaders face a choice that is structurally similar: invest in red flags and try to slow the transition, or invest in road building. The world is changing at a blistering pace. Evidence from every other industry facing this kind of disruption shows that it is inexorable. The firms that will thrive will find the discipline and courage to protect the new from the old, while extracting as much value as possible from the old as it draws down. The clock is not a friend.