The law firm of the near future may be run by people who did not go to law school in order to think about software at all. That is because artificial intelligence is not arriving in Big Law as a discrete product category, like e-discovery or videoconferencing. It is arriving as a medium: an environment that alters the conditions under which legal work is found, understood, produced, checked, and delivered. And, as with every consequential medium, its effects are likely to be felt first in the organization that adopts it, not only in the tasks it automates.
This is not an argument against lawyers. Lawyers possess the thing any serious legal-AI effort needs most: an understanding of legal work as it is actually done, with all its judgment calls, sensitivities, exceptions, and client-specific stakes. But the profession has made a curious institutional choice. Faced with a technological transformation, many firms have built “innovation” functions staffed largely by people with J.D.s.
The assumption is understandable. Legal problems require legal knowledge. Who better to modernize a law firm than someone who understands its work?
But knowledge of the work is not the same as the capability to redesign it. And when the medium changes, preserving the old organizational logic is rarely enough.
A J.D. can confer credibility within a firm, fluency in its language, and a useful instinct for the places where legal judgment cannot be automated away. It does not, by itself, teach product management, systems design, data architecture, user research, delivery operations, organizational change, or the hard craft of taking a promising prototype into the routines of hundreds, or thousands, of professionals.
The result is a familiar innovation structure: well-meaning, intelligent lawyers attempt to become translators, strategists, procurement specialists, project managers, product owners, and occasionally amateur technologists. They are expected to interpret every new development in artificial intelligence, identify commercial opportunities, reassure skeptical partners, understand security and risk, and somehow deliver measurable results. It is an impossible brief disguised as a progressive title.
This is the peculiar danger of treating AI as an add-on to the existing firm. A new technology is acquired; an innovation team is assembled; pilots are launched; the language of transformation begins to circulate. But the structure of work, ownership, and decision-making remains recognizably intact. The firm has introduced the medium without yet accepting that the medium may require a different institution.
These teams can become very good at talking about innovation. They can convene panels, run pilots, circulate surveys, and furnish client pitches with the appropriate notes of technological optimism. Yet too often they operate at a distance from the disciplines that make a product durable: knowing the user well enough to define the real problem; establishing a clear owner; making trade-offs; measuring whether the solution works; and improving it after the announcement has faded.
Product management is sometimes treated as a Silicon Valley import, a set of rituals involving road maps, backlogs, and brightly coloured digital boards. Its deeper proposition is more radical, especially in a partnership: that an organization should begin with a user’s real need and build toward a repeatable outcome, rather than begin with available technology and seek a flattering use for it.
That is not merely a management philosophy. It is a way of recognizing that the medium has changed the client’s expectations. Clients will not care that a firm has access to generative AI. They will care whether the firm can use its new capabilities to make advice clearer, faster, more consistent, more responsive, and more valuable. In other words, they will judge the experience, not the press release.
This is why knowledge management now belongs at the centre of the argument. For years, many firms treated it as a supporting function: a careful custodianship of precedents, templates, and accumulated wisdom. AI has exposed the error in that hierarchy. The quality of a legal-AI system will depend not simply on the model it uses, but on the quality, structure, accessibility, and governance of the knowledge surrounding it.
A firm’s institutional knowledge is not a warehouse of documents. It is a living body of judgment: what has worked before, what must be adapted, what is current, what is client-specific, and what should never be reused without scrutiny. Knowledge professionals understand that retrieval without context is merely a faster way to be wrong.
In an AI-enabled firm, knowledge management is no longer just the preservation of the past. It is part of the production system for the future. It determines what the firm can safely retrieve, what it can reliably reuse, what it can turn into a client-facing capability, and where human judgment must remain visible and accountable.
The firms that make progress will stop treating AI, knowledge management, and product delivery as neighbouring departments competing for a place in the innovation org chart. They will see them as one capability. That means lawyers alongside product managers, knowledge engineers, designers, technologists, data specialists, and operational leaders—not lawyers asked to impersonate all of them.
The question is not whether lawyers belong in legal innovation. Of course they do. The question is why the profession continues to mistake a legal credential for a complete innovation strategy.
Artificial intelligence is forcing law firms to confront what they have long been able to avoid: real working solutions require different kinds of expertise, and expertise cannot be acquired merely by changing a job title. The firms that understand the medium—and allow it to reshape their structures, not just their marketing—will build useful systems for clients.
The others will continue to produce the more familiar product of the innovation economy: an eloquent account of what they intend to become.
