Why is legal tech so hard to get right, and how can we change that with AI?
When I was an early customer of Moveworks, my team and I learned firsthand the importance of specificity and focus when applying AI to an enterprise use case. In the case of selecting Moveworks, we elected to start small by using AI in one problem area — IT support. The product's success led to the autonomous resolution of sometimes up to 40% of support tickets in a month. That small-scale win gave us insight into how similar technologies — automation and autonomous resolution — could be applied across different functions. It was like a starter kit for increasing productivity across the organization.
We quickly discerned that deploying AI within corporate environments is not a destination. It is a journey that should begin with a specific and defined area of focus.
Today, there is plenty of excitement about applying AI to corporate use cases. But the conversation around how to do this in an impactful and effective way hasn’t entirely caught up to that level of hype, mainly when talking about the more complex areas of typical corporate structure. Exhibit A: enterprise law. No function is more sophisticated, nor impacts a broader swath of the average company, than legal.
So, what is the best way to apply AI to the legal world? The topic is still very much up for debate. However, looking at things from the point of view of vendors and buyers provides the framework for a fruitful implementation strategy.
Applying AI to legal: the jury is still out
Scores of legal specializations exist, from commercial and tax law to labor and environmental law to criminal litigation and patent law. The average corporate legal team represents dozens of these specialties and niche functions depending on the company. Regulations and laws change and evolve over time, leading to newer areas of specialization that require different sets of expertise and experience. One that is only now starting is the law around AI.
Simultaneously, law school enrollment fluctuates over time, hitting an all-time valley in 2023. So the human resource component to such an industry is slowly declining when usage of legal services only tends to increase. The workload for legal departments, firms, and individual attorneys has only grown. Even when a company is not doing well, it requires a working legal function for operational purposes. There is a clear-cut need for technologies that can reliably help with some of the heavy lifting. The AI opportunities are vast, from automation to reasoning and negotiation to production and output. From drafting patents to legal search to assessing legal documents for purposes of summarization.
Again, however, we run into the issue of complexity. Who is qualified to create technologies that can service highly specialized legal teams? In the overwhelming majority of cases, only highly trained legal professionals. That hasn’t stopped technologists from trying, of course. Most legal teams use some tools or technology to automate manual work. But the vast majority of these technologies were initially built for another purpose that doesn’t capture the nuance of legal work in a way that is efficient or entirely trustworthy.
The vendor’s perspective
To that end, it is critically important to establish a specific foundational model for the function that combines understanding legal processes AND the law. The tailoring and adapting of an AI model to a particular legal function will ultimately be required from technology vendors. This also means that the relevance of the model must be particular, focused on activities like taking in regulations, interpreting legal opinions, combining that with intelligence documents and internal languages or acronyms, and more. It will require integration with real-time informational sources that update the model with new laws and regulations. In short, it will take an incredible amount of training data to implement this.
First and foremost, the exemplary model architecture must be chosen for the best processing of natural language, one that incorporates outputs and prompts specific to the task at hand. For example, the follow-up prompts would change if you had a ChatGPT-type application for legal teams. Instead of asking if you want more information, it would ask, “Do you want to see precedent cases?” or “Do you want to see standard clauses from your most recent contracts?” Ultimately, contextual understanding is what will define great AI for legal teams.
Systems that understand the level of depth that goes into the function are critically important. The legal world is divided into countless niche disciplines and specialties like medicine. There are massive differences between what is required of courtroom litigators and corporate lawyers alone, and that doesn’t even account for the dozens of specializations within each discipline. Imagine selling the same tool to a public defender and a patent attorney. It would be akin to handing a surgeon’s scalpel to a radiologist. They barely encounter any of the same scenarios; how can one tool effectively serve both?
The conversation should start with applicability and prioritization. What is the most pressing problem a legal team is looking to solve? Does the solution mirror that problem and resonate with the intended customer? If a potential customer is not using external litigators, legal billing and tracking is probably not a critical requirement for any tool they adopt. If they are, it will probably be the highest priority feature.
Recommended by LinkedIn
The CIO’s perspective
From a CIO’s perspective, the same fundamentals apply to evaluating legal tools to any new technology.
First, core basics around security and protection are fundamental. The bar is inherently higher for legal teams than anyone else due to the sensitive nature of the information they’re dealing with daily. Rarely do CIOs encounter a legal-specific tool that doesn't fulfill basic security and data protection requirements. Second, how does the tool integrate into existing workflows? Is it a brand-new tool in the stack, or does it plug into and improve an existing process? Third, how does it integrate into existing architecture? Because legal is one of those rare functions that impact the entire company, it’s critical that the use cases for a tool are very prescriptive in terms of what the solution is and is not designed to do.
Part of the role of the CIO is to ensure that you are continually up-leveling peoples’ productivity, irrespective of their department. Most departments today have a clear-cut digital strategy and specific technological tools that allow them to execute effectively on that strategy. Marketing has a digital strategy. Sales has a digital strategy. Finance has a digital strategy. However, legal teams have always been underserved despite (and perhaps due to) the complexity of their requirements. They are typically left to use either the most essential tools or to make do with tools designed for someone else.
Most other functions also have a pseudo-technology team that helps keep their tools up and running — marketing ops, sales ops, etc. — while legal does not. There is no legal systems admin, so vendors must be extremely clear about enablement, integration, administration, architecture, and workflow.
Plenty more on the docket
Whether it’s reviewing contracts or redlines, or driving a better intake workflow, there is no shortage of pent-up demand and frustration among legal teams. And now, with an entirely new realm of possibilities unlocked by AI, the opportunity to meet that demand has never been more tangible.
Over the next several months, we’ll explore that opportunity in great detail. We will discuss today's specific tools, what they can accomplish, and where they still fall short. We’ll chat with lawyers across many disciplines and those who have already become founders in this space. We’ll speak with GCs and legal leaders in and outside of enterprise tech to get a clearer picture of their challenges and how technology can help them meet them more effectively going forward.
Along the way, I want to hear from you. Are you a founder in the legal tech space? Let’s talk.
Super interesting!
Strategic Leadership, $300M+ Deals | 6 Patents | Client Partner | Sales-Solutions Engineering | ex- IBM, Nextlabs, Arista
2moGreat insights on AI's potential to transform legal operations, Yousuf! The legal sector is indeed ripe for innovation, and your perspective on effective AI implementation is valuable. Looking forward to seeing how AI continues to reshape legal tech and improve efficiency in the field.
Immigration Attorney | AI Tech Founder | Helping Lawyers Scale Sustainably
2moI couldn't agree more! The legal industry is in desperate need of AI automation. Lawyers are quitting at rates higher than most other industries.
For in-house use cases, the trick is something that can integrate seamlessly with other cross-functional teams' preferred platforms (slack, salesforce, jira, etc.) on the front end while being robust and nuanced enough on the backend for legal advice.
I am still experiencing headwinds from internal legal depts and external counsels to more or less fully embrace technologies that are widely adopted by others professions. No collaborative editing, no platform-driven contract management process, etc. are still real, and I humbly think it is partly driven by reticence to adopt newer tricks of the trade.