Large language models

Generative AI on your contract data

Name: Henchman
Price range: $

ChatGPT’s unprecedented growth to 1 million users sparked global interest in Large Language Models (LLMs), which are nothing but astonishing. The legal profession has been particularly invested in exploring the potential of AI and GPT – If you could bill the hours spent by legal professionals debating the impact of AI (and LLMs), you could probably buy Microsoft by now.

If you could bill the hours spent by legal professionals debating GPT, you could probably buy Microsoft by now.

In this memorandum, we will elaborate on how you can safely leverage AI, trained on your own data, to enhance your daily workflow and benefit from your organization’s unique legal context. We will also dive into the rise of what we’d like to call the ‘Augmented Lawyer’ and how assistive smart technologies like Henchman enable ‘Dynamic Knowledge Management’.

The ultimate embedded use of AI in Legal

Plenty has been written about the use of GPT in the legal industry; its pros, cons, and hallucinations. Instead, we want to take you on a journey of logic and concrete examples of how you can make AI and Large Language Models (LLM), such as GPT, work for you in a safe and very powerful way. Legal professionals, knowledge teams, and innovation teams explore how they can have a magical experience such as ChatGPT on their own database. Sounds great right? Welcome aboard.

1: GPT (LLMs) as is, and with a legal twist

1.1: GPT is a language model that is very performant in predicting the next best word with the highest probability. It is great at writing expressive sentences and following linguistic rules, but cannot really understand WHAT it is writing, hence its inherent hallucinations or inconsistencies.

1.2: But let’s say we ONLY train the Large Language Model (LLM) on legal information, documents, and contracts. That would make a lot more sense right? Open AI-based solution, Harvey, is a great recent example of such an approach, where the technology behind GPT is trained on public legal data.

2: GPT + YOUR data

Connecting your entire database to AI like GPT seems like a cure-all, but it falls short as well of being a good solution. First of all, you have to be comfortable with using your data in the system of a third-party vendor (often with specific regional data compliances). Secondly, while the results might mimic your own clauses a bit better in some cases, you still never really know what historical data was used.

Let’s take another concrete contracting example where an associate prompts your data-induced LLM with the following:

We hear you think “Excellent”!

Hold on … “Is this a clause we often use in the firm?” “Did it take the wording from a signed contract or an earlier version?” “Is this a clause my senior partner would approve?” “Has this clause been redlined by a client before, and why?” …

When drafting and negotiating complex contracts *these* are the answers that would make or break the right decisions for any client.

This data-induced LLM remains a generative language model that is very good at predicting the best possible next word, but still lacks any form of data insight or legal context.

3: The Holy AI Trinity

3.1: An intelligent layer on top of YOUR data:

When drafting contracts of a more complex nature, you require the best of both AI and data with a better awareness of the legal context and the context of your office.

An increasing amount of law firms are implementing an intelligent layer, like Henchman, that sits on top of the firm’s Document Management System (DMS) or contract database.

This technology connects to the legal team’s database and processes all clauses and definitions automatically (a number that can easily approach a couple of millions) – without any manual effort required! Well-known contract databases are iManage, NetDocuments, Sharepoint, Google Drive, etc. During this process, it links all previously existing metadata to the clauses, applies automated tags, such as “from signed contracts” and applicable law, and adds statistical data on how common a clause is within your database. All this information gives the drafter all the relevant context when drafting and negotiating contracts.

3.2: An intelligent layer on top of YOUR data, ENRICHED with GPT

Bring those three elements and witness complex drafting on steroids:

Have your legal Document Management System (DMS) content enriched using AI, so you can easily organize it according to complex or interconnected criteria
Respond to much deeper knowledge questions than a traditional AI + Data system combination, such as deep questions about the commonality or the past history of clauses within their contract contexts

Some examples:

What is a common claim period and liability cap in this particular case?
Give me a list of periods that were commonly used for this type of contract and what differentiates their use.
Give me similar contracts to use as inspiration, and summarize the differences between these contracts.
I’m writing a complex purchase agreement, how buyer or seller-friendly is my construction?
Can I get a list of buyer-friendly clauses to balance out this agreement?
This clause seems to have a precedence, but it is very uncommon – is there a more common replacement?

Dynamic Knowledge Management

This data-driven approach to drafting and the deep understanding of a firm’s database is opening doors to think of how we gather, maintain, and distribute ‘knowledge’ (think: templates). Legal professionals are increasingly aware that we need a new way to make knowledge more accessible. A way that is effortless to create, maintain, and distribute. Something we like to call Dynamic Knowledge Management.

At its core, Dynamic Knowledge Management is about having the right insights, at the right time and in the right context. It’s about making sure that the time spent by everybody in the legal firm is spent efficiently and it’s about having a way to dynamically curate knowledge throughout the firm. Contrary to the traditional version, dynamic knowledge management is a continuous process of finding trends in the data set and serving the right piece of curated knowledge to the lawyer working on a new case with the same context.

Here is an example of Dynamic Knowledge Management in full force:

Let’s say a law firm is increasingly assisting PE firms with the purchase of solar panel companies. They have developed tailored SPAs and expertise for these types of transactions. The dynamic knowledge management system has the ability to detect this trend and distill the nuances in contracts, clauses, and definitions for this type of transaction.
When it’s time to start, draft, or review an SPA for another solar panel company transaction, the platform can show the most appropriate curated knowledge for that context (e.g. sector). This curated knowledge is carefully matched with and applied to the contract or clause the lawyer is working on.
All of this happens automatically and without manual effort prior to or during the contract drafting process by the lawyer or knowledge management team.

For lawyers, it means being able to easily find and compare similar cases, being confident in making decisions in their specific context, and applying changes backed by nudged golden standards and insights from previous cases.

For knowledge managers, it means efficiently allocating time and effort by identifying and updating the most important and relevant content, based on trends and data. Enabling teams more efficiently by sharing relevant information and resources where and when the drafter needs it.

With Dynamic Knowledge Management, law firms will effortlessly transform in real-time their raw data and accrued knowledge into valuable assets that can help them stay competitive in today’s fast-paced legal landscape.

Knowledge Management + Innovation Management => Legal Enablement

Innovation managers play a crucial role in driving digitalization across various industries. They are typically brought on to address a specific organizational need and move on once the task has been completed. However, their hunger for innovation often leads to the creation of new tools, processes, and even entire job categories that can drive efficiency and solve global problems.

This trend can be seen in industries such as customer success and marketing automation, where innovation-driven approaches have led to the creation of millions of jobs and billion-dollar companies. In the legal industry, we are also seeing the emergence of innovation managers who are tasked with centralizing knowledge management and facilitating continuous education for legal professionals. This could potentially lead to a new role called Legal Enablement, which could greatly benefit legal teams and clients alike. With the right solution for dynamic knowledge management, legal teams can start making progress today.

In conclusion

GPT and other Large Language Models are truly remarkable technological innovations that are poised to revolutionize the work of legal professionals. These models are incredibly versatile and can be used for a range of tasks, such as generating summaries, contracts, and clauses, providing valuable starting points or sources of inspiration. When integrated with a Document Management System, these models can produce text that reflects the unique tone and style of a firm.

It is worth noting, however, that while GPT and other generative technologies are excellent at predicting the probability of the next word, they lack context. In order to ensure that the text generated is accurate and appropriate for the specific context of a document, it is important to incorporate an intelligent layer that includes metadata such as client name, contract type, author, version, and commonality. By doing so, legal professionals can have greater confidence in the output generated by these models, particularly when working with complex contracts that require a nuanced understanding of the legal landscape.

This Henchman article has also been published by Law.com

Henchman customers