Key Takeaway: Every law firm now has access to powerful AI, but that advantage is disappearing fast. Models are becoming commodities, and the competitive advantage for firms lies in the delivery of trusted, governed context to those tools. The time to build that foundation is now, before competitors turn knowledge into lasting differentiation.

Working with firms for the past few years, my Epiq colleagues and I have been party to all stages of AI adoption across large law firms. We have seen firms experiment, move past pilots, and deploy tools across their practices. Now the availability of AI platforms and tools is ubiquitous, with firms employing at least one if not multiple leading AI tools.

In this sense, every firm is not equal. Adoption depth depends on the team, as do workflow redesign and business model updates. Those differences are real, but they can be copied. A firm that reworks its diligence process well will find competitors have done the same within a year or two. The one advantage that compounds, and that no competitor can license or replicate, is the context a firm brings to the tools it uses.

The Model Is a Utility. The Context Is Not.

If every firm uses the same models, the model itself is something anyone can license. The real question is not which model a firm chooses, but what it gives that model to work with. When grounded in a firm’s matter histories, precedents, and governed data, with provenance preserved throughout, the output becomes more than an assertion; it becomes something a partner can trace to its source and verify. The law of “garbage-in, garbage-out" remains. Firms must control what goes in and when, and how it is used to truly deliver a high-quality differentiated service. 

Outside counsel guidelines (OCGs) and ethical walls create unique constraints for firms in providing context to AI models. Any serious approach to providing context must work within those constraints rather than around them, and this is where the difficulty lies. 

Knowledge Management Now Serves Two Constituencies

Traditional knowledge management (KM) exists to enable people to identify knowledge. In an AI firm it acquires a second constituency of equal importance. AI systems must also be able to locate and apply what the firm knows. KM, understood this way, is the discipline of bringing the right context to intelligence exactly when they need it.

Meeting that standard requires more than a precedent chatbot. It requires an operational foundation with several connected components.

1. Knowledge captures institutional experience, i.e., the judgments and reasoning behind the documents. Without it, a firm has a document store rather than a knowledge base.
2. An ontology establishes shared meaning so that a term used in one practice group maps to the same concept in another. Without it, retrieval returns volume rather than relevance.
3. A retrieval layer surfaces the right context at query time, with permissions and citations enforced. Fail to implement, and the confidentiality problem above becomes unmanageable.
4. Skills and playbooks codify how the work is performed, not how it is described in a template. Without them, the output is generic.
5. Workflows put the whole practice to use. Otherwise, the foundation exists and nobody touches it.

Together, these make up context infrastructure: the systems, structures, and operating models that turn institutional knowledge into AI context.

Own the Context, Rent the AI

Tools, models, and platforms are changing quickly enough that today's best option may be behind the field within a year. That argues against depending on any single provider. Firms are better served owning the context infrastructure and renting the AI. Take up the strongest available tools as the market progresses and surpass those that fall behind without disturbing the foundation underneath.

Ownership needs definition, though, because the binary is not clean. Most firms' matter data already sits in a document management system, and those suppliers are building context layers of their own. In practice, a firm will rent the models, and probably the document store. What it should own is the semantic model, the ontology, and the mapping between them. That is where firm-specific meaning lives, and that is what is most expensive to rebuild elsewhere.

Providers will change. The context should not.

The Questions Worth Asking

The work begins with a candid assessment: 

• What context assets does the firm hold today, and where do they sit? 
• Which assets are genuinely strategic rather than merely large? 
• Who owns them, and who governs access when a matter closes or a team is walled off? 
• Are they in a form an AI system can retrieve and reason over, with permission and provenance intact?

Every major firm has access to powerful AI. Not every firm will have a differentiated context infrastructure, and it is in that gap that the near future of competitive advantage will be decided.

Find out where your firm stands with Epiq Law Firm Advisory.


Eric Anderson, Senior Director, Technology, Data, and AI
Eric brings more than 20 years of experience at the intersection of business, technology, and law to his role at Epiq. He builds enterprise data platforms and data lakes using Azure Synapse and Microsoft Fabric. His work enables law firms to adopt data-driven strategies and apply analytics to legal workflows.