Waiting for the Legal Revolution

Legal Tech for many has been a field of interest for a long time. And, although there have been interesting things happening here and there, we so far have not seen the large scale transformation of the legal sector, so many are eager for. In this piece I dive into the reasons behind the slow movement of legal tech and explain what is different this time and why now is the best time there ever was to get into legal tech.

Doing justice justice.

First, let me give you a clearer picture of what Legal Tech is and why it is important. Legal Tech (i.e. the application of technology to legal workflows) seeks to address pressing societal challenges within the traditional legal landscape. Chief among these are barriers to accessing justice, often due to the high costs of legal services or a lack of awareness about rights. The existing system also grapples with inefficiencies, leading to time-consuming processes that can exacerbate hardships for those involved. Compounded by the overwhelming volume of legal information, the opacity of legal procedures, and geographical or physical constraints, many individuals find it difficult to navigate the legal world. Furthermore, inconsistencies in the delivery and quality of legal services, coupled with a general lack of legal education for the public, further underscore the need for innovative solutions. Legal tech embodies the promise to democratize, streamline, and enhance the realm of legal services for both professionals and the broader public. Due to the foundational role our legal system plays for society and the manifold pain points we think that Legal Tech represents a major opportunity for AI to do good in the world.

However, there is lot of work do be done. In 2021 the estimated size of the legal service industry was 900$ bn, while the legal tech industry was only worth an estimated amount of 24$ bn. This difference provides a hint on how far tech penetration in this particular industry is lagging behind. To understand the current state of legal tech and how, AI can help with the challenges described above, we talked to Viktor von Essen, accomplished lawyer, legal entrepreneur and founder in residence at Merantix. (You can listen to an excerpt of our interview here, or scroll further down to read up on the highlights: 1×0:00-11:56)

Legal Tech - the current state and how we got there

The phenomenon of legal tech (i.e. applying technology to legal workflows) is nothing new. The term has been around since the 1970s, when the first legal databases were developed. However, so far, there has been no substantial change to the legal services industry. To understand, why most of the attempts failed and what has changed in recent times, we need to take a look at the newest history of the field. If we look at the past ten years of legal tech, we can roughly differentiate between three phases.

• pre-pandemic: Before the pandemic, the legal tech industry's approach was predominantly centered on rule-based systems and no/low code solutions. Rule-based systems, a vestige of traditional artificial intelligence approaches, relied on fixed algorithms and predefined scripts. These systems required explicit programming for every possible scenario they might encounter, making them rigid and less adaptable to the complex and nuanced world of legalities. No/low code platforms aimed at simplifying the creation of applications by reducing manual coding. In the context of legal tech, this meant attempting to streamline litigation processes, making them more accessible and user-friendly. While the promise was high, the delivery often fell short due to the intricate nature of law, which frequently requires bespoke solutions that generic platforms struggled to provide. Machine Learning (ML) models as a potential solution presented its own set of challenges. The amount of data neccessary and the costs of training your own models made many applications unrealistic. Many promising trends and budding technologies seemed to hit a "glass ceiling," where further progress appeared elusive. In essence, the pre-pandemic period was one of high hopes but limited tangible advancements in the intersection of law and technology.

• post-pandemic: The pandemic, with its wide-reaching implications, catalyzed a seismic shift in many industries. For the legal realm, it ushered in an accelerated wave of digitization, the magnitude of which was both sudden and transformative.

  Take dispute resolution, for example. Traditionally, this process necessitated physical presence, with courts and mediation rooms serving as hallowed grounds of justice. But as the pandemic imposed unprecedented restrictions on movement and gathering, the industry was compelled to adapt. Enter digital hearings. Almost overnight, courtrooms and mediation sessions transitioned to platforms like Zoom. This wasn't just a temporary adaptation but marked a fundamental shift in how justice could be both sought and delivered.

  Beyond hearings, other facets of the legal process experienced digital transformation. The long-established practice of affixing 'wet' signatures on legal documents, a practice often deemed sacrosanct, began to wane as digital signatures gained acceptance. This might seem like a small change, but its implications are profound. The acceptance of digital signatures not only streamlined processes but also symbolized the legal industry's willingness to evolve and embrace modern solutions. This new spirit was best encapsulated by Joshua Browder, who remarked, "I think, COVID has accelerated the legal industry by 20 years!" It wasn't mere hyperbole. The inertia that had long characterized this traditional industry was shattered. The base layers established during this period are not mere stopgap solutions but foundational changes upon which future innovations can be built.

  While the pandemic's toll has been steep, in the context of legal tech, it has unlocked doors and paved avenues for progress. The legal community's increased willingness to adopt technological solutions post-pandemic signals a brighter, more efficient future for both practitioners and those seeking justice.

• post-LLM/now: The advancement of Large Language Models (LLMs) represents a paradigm shift in the realm of legal tech, ushering in a transformative era that promises to redefine the very fabric of the industry. In the last months we have seen GPT-4 passing the Bar exam and AI outperforming humans on tests in reading comprehension and language understanding. While legal professionals have always been aware of the winds of change, the magnitude and potential of LLMs have truly captured the industry's collective imagination.

  Historically, law firms have been bastions of conservatism. Yet, the promise of LLMs has galvanized even the most traditional firms into action, signaling an unprecedented willingness to embrace digitization and automation. Examples for this are Thomson Reuters, acquiring Casetext for $650m or Law firms, such as Dentons, developing proprietary LLMs in house. This isn't merely about jumping on the bandwagon; the results are tangible. Consider the realm of Q&A solutions: AI-driven platforms, bolstered by LLMs, are now able to handle an estimated 60% of tasks that, just a year ago, were the exclusive domain of human practitioners in areas like litigation. The law, at its core, is a profession rooted in language. Legal documents, contracts, statutes, and courtroom arguments are all expressions of language, and this makes the rise of LLMs particularly pivotal. Models, which excel at understanding, generating, and manipulating language, align seamlessly with the legal profession's needs. However, the journey has only just begun. While the current LLMs have been trained on vast swathes of the internet, capturing the essence of general language, they haven't been tailored specifically for legal workflows. Despite this, their applicability is evident, showcasing their inherent versatility. But the real revolution lies ahead. As the industry begins to create specialized legal language models, trained and fine-tuned for specific legal tasks and jurisdictions, the true potential of LLMs in legal tech will be unveiled.

• outlook on the near future: Traditional law firms, structured around hourly billing, often have incentives to maintain the status quo. However, as technology offers pathways to efficiency, these old-guard entities are bound to face pressure. New players, equipped with cutting-edge technology and innovative business models, are poised to challenge the established order, placing incumbent firms under increasing strain. This shift is not just about the firms but also about redistributing power within the legal ecosystem. To illustrate, consider the dynamics of contract drafting: A corporate legal counsel needing a labor law contract might currently send it to an external specialist, incurring significant costs for a mere endorsement. With the advent of sophisticated LLMs, this dynamic is set to change. The in-house lawyer could soon draft or refine contracts directly, with external counsel acting more as an escalation point rather than a primary drafter. This democratizes access to legal services and redistributes power back to those seeking legal advice. In the immediate future, a key milestone will be the development and deployment of legal language models tailored for specific European jurisdictions, reflecting the unique legal landscapes of each member state. In the coming years, we can anticipate a proliferation of these specialized models, each serving a unique purpose within the legal framework. From contract analysis to predictive judgments, the horizon is vast and full of promise. The dawn of the LLM era in legal tech isn't just another phase; it's a transformative moment that promises to redefine the industry's future.

Applying AI

We've already grazed the surface of AI's potential in Legal Tech, like its role in Q&A systems. But let's dive deeper and categorize these use cases for a clearer understanding:

• Drafting Legal Documents: This could range from simple contracts to more complex legal documents. AI can assist by standardising documents, ensuring they adhere to relevant laws, and making suggestions to improve clarity.

• Extracting Information from Legal Documents: AI can swiftly sift through mountains of data to extract relevant information, something that would be time-consuming for a human.

• Asking Questions (Legal Q&A): Whether it's a basic legal query or more intricate matters, AI can be at the forefront to provide instant answers. This could potentially be a subset of extracting information since it revolves around pulling relevant details from legal repositories.

• Summarisation: Legal documents are notorious for their length. AI can distill these documents into concise summaries, preserving the core information.

This just reiterates the transformative potential of LLMs in reshaping the legal landscape.

Specific Use Cases:

1. Simplifying Legal Complexities for Consumers/Clients

Imagine getting an eviction notice. Instead of running to a lawyer immediately, you could consult an AI-powered chatbot to guide you through the legal intricacies. A caveat, though: Legal regulations vary, and while AI can provide information, the line between "legal information" and "legal advice" is thin. Given recent shifts in legal markets, like in Germany, there could be more room for AI in the future.

Examples:

• DoNotPay: AI to help users fight big corporations, protect privacy, find hidden money, and beat bureaucracy. (backed by a16z and COATUE)

• rightmart: Vertically integrated platform for providing legal services from law firms to customers. (backed by LVM Versicherung and KS/AUXILIA)

2. Legal Workflow Optimisation

This might seem like an expansive category, but it holds merit. For instance, when lawyers analyze cases, determining the applicability of specific laws can be automated, making the entire process more streamlined and efficient.

Examples:

• Ironclad: Contract lifecycle management platform used by companies to handle every type of contract workflow. (backed by Accel and SEQUOIA)

• Luminance: Document review empowered by supervised and unsupervised machine learning. (backed by Talis and SLAUGHTER AND MAY)

3. Digital Judiciary

Envisioning digital courts and e-judgements isn't sci-fi. But such advancements might necessitate legislative changes to accommodate and regulate them.

4. Automated Dispute Resolutions

Mediations and resolutions don't always require a human touch. Platforms like eBay already employ AI to partially auto-resolve customer complaints, especially in B2C contexts like eCommerce.

Examples:

• EvenUp: Empowering personal injury and victims lawyers using AI. (backed by Bessemer and BainCapital)

In all these scenarios, the intersection of AI and law can lead to more efficient, streamlined, and accessible legal processes. The coming years will undoubtedly witness the evolution of Legal Tech, with AI at its core.

Current challenges for legal AI

Privacy and Trust

Challenge:
The management of sensitive data is a significant concern when integrating Large Language Models into legal workflows. Moreover, the high-stakes nature of legal problems demands a high level of trust in the technologies employed. LLMs are not yet entirely reliable and can generate incorrect or "hallucinatory" data.

Solution:
Emerging techniques like open-source models that are locally hosted, fragmented learning, and privacy-preserving technologies are being deployed to handle data sensitivity. Additionally, building explainability into these models and maintaining a human-in-the-loop can validate the AI's output, thus increasing trustworthiness.

Precision and Accuracy

Challenge:
Legal documents and procedures require an extremely high level of precision. Any slight mistake or inaccuracy can have severe implications, making the margin for error almost nonexistent.

Solution:
Well-trained and fine-tuned LLMs can potentially be more accurate than humans for certain tasks. The human-in-the-loop approach will continue to be essential for oversight, making sure that the language models' outputs meet the rigorous standards of legal accuracy.

In summary, while some challenges remain in incorporating AI into the legal industry, solutions are actively being developed. The field is on the verge of transformative changes that will likely make legal processes more efficient, more accessible, and more accurate.

Thanks for reading, and please let me know what you think, what further opportunities in the market you see or what technological trends on the horizon you believe will make an impact. A special thanks goes out to Eduard Hübner, who was my great co-author for this piece.