By Charles Brecque, CEO and Founder of Legislate.
Data and artificial intelligence are transforming the legal technology space—there’s no doubt about it. A recent Thomson Reuters Institute survey of lawyers showed that a large majority (82%) of respondents believe ChatGPT and generative AI can be readily applied to legal work.
While it’s tempting to think of legal tech as a playground exclusive to law firms, as technology enables employees without legal training to use and create legal frameworks and documentation, I’d like to challenge that narrative. Being the founder of a company that uses AI to manage contracts, the way I see it is the real magic happens when legal tech tools meet the day-to-day challenges of small and medium-sized businesses (known as “SMBs”).
But why SMBs?
At first glance, law and SMBs might seem disconnected, but at the heart of every business are contracts. Contracts define collaborations, seal deals and are the backbone of commerce. Yet, for many SMBs, managing these contracts with suppliers or customers can be complex and tedious, especially with limited legal resources, as most SMBs can’t afford to invest in big legal departments.
Enter large language models as applied to legal tech. With LLMs, we’re not just talking about a smart tool that can decode legal jargon, though that’s certainly a part of it. We’re talking about a legal copilot that can generate, analyze search and optimize contracts in real-time. Beyond conserving time and resources, LLMs can reveal intricate patterns in contracts, highlight specific traits in the most beneficial agreements, pinpoint contracts that endure prolonged negotiations and discern terms that regularly face challenges or neglect.
By tapping into the capabilities of LLMs, SMBs can not only transform their approach to contract management but also refine business decisions related to their sales efforts.
Nothing good comes without challenges.
Like all good things, bringing LLMs into SMB legal tech doesn’t happen without its challenges. For businesses to successfully implement and put their own models to use for contract management purposes, there are a few considerations to take into account.
Solving The Hosting Hurdle
LLMs demand a lot in terms of infrastructure. One question for SMBs is whether to lean on third-party cloud solutions or invest in their own machinery. Data privacy, coupled with the costs of hosting models on graphics processing units, makes this a critical decision point. Here, solutions provided by companies that offer a way to compress LLMs can help improve how businesses implement, maintain and manage them.
Navigating Integration And Usability
Integrating LLMs into SMBs’ contract management workflows has consequences, particularly in terms of the employee adoption of these models. Leaders need to understand how the LLMs will be used across the organization—be it by the legal team or another department. The experience and background of these users will determine whether they can judge LLMs’ output appropriately.
Choosing a legal tool or provider that doesn’t offer excessive freedom in prompt selection is one way to help mitigate potential errors or “hallucinations.” If, however, you’re developing your own model, it’s essential to ensure that users are trained to recognize unreliable outputs. A useful tip is to always ask the LLM for the source of its response. If it can’t provide one or cites an incorrect source, it’s likely hallucinating.
Second, SMBs need to think about LLMs’ compatibility with their current technology and their in-house expertise—and factor in the time to adapt. Each LLM can have distinct architectural requirements. While most software operates on central processing units (or “CPUs), LLMs typically need GPUs. These not only come at a higher cost but also function differently from CPUs. As a result, IT teams might face a learning curve if they’re not familiar with GPU-based tools.
Finally, it’ll be important to bear in mind the impact on the SMBs’ various stakeholders. There might be varied responses to the adoption of LLMs, and transparency and clear communication are key. Highlighting the benefits of the new tech while also addressing potential concerns ensures all parties remain aligned and comfortable. SMBs can do this easily by creating a code of conduct or LLM guidelines that are openly shared with their stakeholders.
Understanding The Data Privacy Trade-Off
As LLMs develop, they constantly refine their models using incoming data. For SMBs that hold personal or proprietary information in contracts and other legal documents, this is an important decision. Do you provide the LLM with data to boost its efficiency, potentially risking IP exposure to third parties? If you do and you’re working with a legal tech provider, ask whether it operates an LLM on its own secure servers. Similarly, if you work with an LLM provider, ask whether it has an enterprise package that prioritizes enhanced privacy safeguards.
How can SMBs effectively add LLMs to their legal tech arsenal?
For those SMBs looking to adopt LLMs, here are three tips.
1. Identify the use case. Start with clarity. Determine precisely where and how an LLM can benefit your operations. It’s not about having the tech for tech’s sake but about leveraging it for tangible outcomes.
2. Reflect on your data stance. If you’re not comfortable sharing your data on external clouds, look inward. Building in-house might be more intensive initially, but it ensures control.
3. Anticipate the change. Make sure you’re aware of what this adoption will mean for your company, employees and stakeholders. Put together a roadmap and internal guidelines before you activate it to ensure a smooth rollout.
It’s an exciting era for SMBs. While the initial hurdles in adopting LLMs might put leaders off, the long-term benefits are well worth the effort. This isn’t about jumping onto the latest tech bandwagon. It’s about understanding the specific needs of the business, appreciating the potential pitfalls and leveraging technology for meaningful, tangible outcomes. As with any tool, it’s not about possession but application.