Law Firms Are About to Start Systematizing Legal Judgment — And the Implications Go Far Beyond AI
Legora just officially announced the launch of Skills, and the timing is good, because legal AI skills have been on my mind for months.
When I heard the feature was being built a couple of months ago, I volunteered our Strategy and Innovation team at Stewart McKelvey and The Fortified Firm for the early test group, and we got beta access in early July. I spent a few days on vacation diving deep into it (sad perhaps, but this is fun for me!), partly to prepare for my upcoming panel at the Canadian Legal Summit next week, “Getting Reliable Results from Legal AI: Prompts, Workflows, and Emerging Agents,” and partly because I was curious what it could mean for us and our clients as we look to systematize our workflows.
What Are Legal AI Skills?
That work has had me thinking hard about agents, skills, and their equivalents across different platforms for law firms. Some tools call them skills. Others call them agents, plug-ins, playbooks, context, system prompts, or workflows. The labels will probably keep shifting.
What matters is the function: Legal AI skills are the layer where lawyers tell AI how work should actually be done, and where the boundaries, standards, and legal judgment that should guide it get written down.
I don’t think law firms yet appreciate how central legal AI skills are going to become.
Legal Knowledge Isn’t the Same as Legal Judgment
General-purpose AI models, trained on public data and legal databases, already know a remarkable amount about the law. But knowing a lot about something is different from applying it well. An experienced lawyer knows which facts matter and which are noise and which precedents are genuinely useful and which should be ignored. They know whether a client wants the safest available position or is willing to accept the risk, what needs to go to a partner, and what should never go to AI without careful review. Much of what separates good legal work from excellent legal work lives in those judgment calls, not in the underlying legal knowledge itself.
The Tacit Knowledge Problem Law Firms Have Never Solved
Firms have tried to capture this for years through precedents, practice manuals, checklists, mentorship, and knowledge management. All of it has value. But it has always run into the same wall: the most valuable knowledge in a firm is tacit. It lives in people’s heads. It’s the pattern recognition built from hundreds of similar matters. Capturing that has always been slow and expensive. Someone has to sit with the expert, pull out what they know, turn it into a process, document it, train people on it, and keep it current. Even if they could do that, scaling that expertise was exceptionally difficult.
Tools like Legora’s legal AI skills change that calculus. Plain language can now become part of a firm’s operating system. A senior lawyer no longer needs to translate all of that judgment themselves into a rigid decision tree or piece of software. They can explain how they think, in their own words or through their own work, and AI can help turn that into structured instructions that can be tested and reused.
That takes us well past the traditional precedent bank. The useful knowledge is no longer just “here’s the precedent we use.” It becomes a record of when to use it and when not to, which provisions worry us and why, what we ask the client before proceeding, what an experienced lawyer would flag, and the point where senior lawyers must step in. Over time, a firm’s IP will come to include not just its documents and know-how, but this distilled version of its lawyers’ judgment, built into the systems people and AI use together.
Legal AI skills aren’t only about making AI more useful. They’re one of the few real mechanisms firms have for keeping AI in check. We shouldn’t be handing a capable general-purpose model a legal task and letting it decide on its own how that work should be done. As models get more capable, that human-defined layer becomes more important, not less, especially in a field like law, where precedent shouldn’t be applied mechanically and two similar-looking situations can call for different advice. That’s why simply giving every lawyer a powerful model will never be the whole answer. As the underlying technology becomes commoditized, the advantage shifts to what’s built around it: what the firm has taught the model, what judgment it has captured and what guardrails it has put in place.
Who Owns a Firm’s Captured Judgment?
This also surfaces problems firms haven’t had to solve before. If a partner spends thirty years building expertise in an area, and that expertise can now be distilled into something that improves hundreds of other lawyers’ work, how should that be valued? Firms have traditionally rewarded origination, client relationships, and billable hours, not the time spent codifying judgment so it can be reused at scale. Who owns the resulting asset, how is it protected if someone leaves, and who keeps it current as the law changes? And if expert judgment becomes reusable at scale, what happens to leverage, to the economics of work, or to the value of an hour spent refining a system that then shapes thousands of hours downstream? None of this is really a technology question. It touches core capabilities like lawyer development, compensation, pricing, profitability, IP, risk, and firm strategy all at once.
What This Means for Law Firms
I think this points to a genuinely new role inside firms, broader than traditional knowledge management. Firms will need people who sit with practicing lawyers, not just to document what they know but to understand how they think. They will need to turn that into AI-enabled workflows and test them on real matters. Those feedback loops then unlock discoveries for the next iteration. This is not a one-off documentation project but an ongoing capability. This demands an unusual mix of skills: legal fluency to be credible with lawyers, process sense to break down how the work happens, AI literacy to know what can be encoded, and change management and incentives to make sure people use what gets built.
I don’t think firms get to treat this as a choice. If it becomes possible to capture expert judgment and deploy it at scale, someone will do it — whether that’s a competitor, a client building it in-house, or a new entrant built around exactly this idea. The question isn’t whether legal expertise gets systematized. The question is who does it first, who does it well, and who builds the discipline to keep improving it.
In a legal industry where access to the best AI becomes table stakes, differentiation will increasingly come from the firms that develop the capability to systematize their workflows and continuously adapt them as work evolves.
AI may make expertise far easier to scale. But people still have to decide which expertise is worth scaling, and how.
About the author
