Large Language Models Can Drive a SaaS Shift for Big Law Firms

The way law firms deliver legal services is transforming. As firms increasingly use artificial intelligence, the traditional billable-hour model appears more misaligned with AI’s efficiencies.

Flat-fee billing is one response, but a more radical solution could emerge: a software-as-a-service model in which law firms develop custom large language models and license them to clients and other firms. Though the idea of law firms becoming software providers may seem far-fetched, current market and technology trends suggest it’s a realistic possibility.

Across Big Law, law firms are hiring dedicated AI experts and leadership and establishing dedicated AI practice groups. Leading firms such as Dentons and Gunderson Dettmer are even hiring both legal and tech experts to build in-house solutions. Other large companies such as JP Morgan Chase & Co., Deloitte, and McKinsey & Co. have already rolled out custom chatbots for in-house use.

Second, legal AI companies are rapidly improving their AI products by training their foundational LLM models on legal-specific data. This data drives creation of higher-quality, legal-specific LLMs that can deliver better legal services. Top legal AI companies such as Harvey and Leya are hiring former Big Law attorneys as researchers and engineers to create new legal data sets to improve the performance of foundational LLMs.

While legal AI companies are building new data sets, the most valuable data already exists within Big Law firms. Such firms sit on legal documents, templates, case law research, and legal analyses spanning decades and created by great legal minds. Due to attorney-client privilege and other ethical restrictions, Big Law firms are limited in their ability to license or sell this data—only they can access and leverage it for LLM development.

By fine-tuning foundational models on proprietary data sets, Big Law firms could create custom LLMs that reflect their unique style, structure, and best practices. To achieve this, firms may either bring in legal AI companies as contractors—ensuring all ethical requirements are met—or hire in-house developers and researchers to build firm-specific LLMs.

The ability to capture the processes and intangible practices of top lawyers is another edge Big Law firms have in creating custom legal-specific LLMs. To achieve fully operational legal intelligence—the point at which an LLM can perform legal tasks with the proficiency of a lawyer—legal-specific models require more than legal documents. Fully operational legal intelligence requires the documentation of intangible processes that experienced lawyers rely on daily.

For example, a partner’s approach to complex issues—iterating on strategy, allocating tasks, guiding associates, and refining drafts—embodies nuanced judgment and expertise that isn’t reducible to a text-based medium that can be fed to an LLM. To unlock AI’s full potential in legal practice, firms must convert these intangible processes and workflows into structured data, allowing their firm specific-LLMs to acquire high-level decision-making capabilities. Law firms are uniquely positioned to convert the thinking and processes of top lawyers into structured, LLM-trainable mediums.

However, opportune positioning alone won’t be enough to propel Big Law firms toward a SaaS model. Instead, the shift likely will be driven by economic pressures stemming from AI’s enhanced efficiency, the move away from billable hours, emerging revenue opportunities, client-driven pressure, and even state bar ethical guidelines on client data protection and privacy. With top financial and consulting companies—which employ Big Law firms—already deploying custom AI chatbots in-house, services-by-LLM-chatbot may become the norm.

Once law firms establish reliable, custom LLMs for internal use, the next revenue-maximizing step is external licensing. By offering custom LLMs to in-house counsel and even other firms, Big Law firms could create a SaaS revenue model.

In-house teams and firms would benefit from consistent access to high-quality legal support, while firms would expand their reach and strengthen client relationships. Firms might experiment with tiered subscription plans, where clients pay different rates based on their access to certain types of AI-driven legal services or levels of human oversight.

The potential of firm-specific LLMs extends beyond reactive legal services. Integrated within clients’ operations, these models could provide proactive monitoring, identifying potential legal issues before they arise—similar to cybersecurity monitoring software.

In the future, Big Law firms may compete to showcase the quality and precision of their proprietary LLMs, positioning them as core differentiators in the legal market. Firms could use AI performance metrics such as accuracy, capabilities, and delivery speed in client proposals.

But a SaaS law firm model is still years away due to technological, and potentially ethical, limitations.

Fully operational legal intelligence isn’t yet achievable because current LLMs can’t conduct autonomous legal research, drafting, or document review with the precision required of a lawyer. But rapid improvements for reasoning tasks and recent advancements in cross-platform “agentic” functionality of LLMs suggest law firms eventually will overcome technical obstacles to a SaaS-model shift.

Still, ethical impediments may arise. For instance, is it ethical for law firms to train their internal LLM models on past client work product?

As law firms contend with AI-driven efficiencies and a potential decline in billable hours, their future may increasingly rely on a SaaS model. Firms that embrace this shift should begin the groundwork now: cleaning and retaining data, documenting processes, addressing ethical considerations, and training LLMs.

If law firms aren’t the first movers, individual lawyers might take the lead. We may see a future where lawyers develop their own personal LLMs that capture and mirror their unique knowledge, experiences, and workflows, potentially licensing these models directly to firms and clients.

Today, software supports the delivery of legal services; tomorrow, it may become the delivery of legal services itself.

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law and Bloomberg Tax, or its owners.

Author Information

Oliver Roberts is co-head of Holtzman Vogel’s AI practice group at and CEO and cofounder of Wikard, a legal AI technology firm.

Write for Us: Author Guidelines