Whenever there is a significant shift in the industry, we are interested in tracking its implications.
Although many companies have been using AI in legal in some form or other for years now, the advent of ChatGPT and large language models (LLM) that are powerful enough to understand and generate meaningful responses to complex questions without additional training represents a new era. There remain sceptics who don’t believe the technology is advanced enough to have a significant impact in the legal industry. These sceptics find it hard to imagine concrete use cases for the technology in legal practice. There are others who believe LLMs and advanced generative AI will radically change the industry in the space of just a few years.
The truth probably lies somewhere in between.
Having read many of the debates raging online, the difference between the skeptics and the enthusiasts appears to be more than just a personality trait, like the difference between pessimists and optimists. It is rooted in the way that different people imagine the technology being used (from fully replacing humans in a legal workflow, on the one hand, to supporting part of a legal workflow with humans in the mix, on the other). There are also clear differences in enthusiasm between those who have not had first-hand access to LLMs beyond the ChatGPT app, and those who have seen and accessed the extremely powerful beyond-GPT-3 LLMs that some companies and firms are working with in beta. The fact that it is the latter group who tend to be enthusiasts should be telling.
Regardless of how you feel about the tech, there are some real things happening in the industry with LLMs. I started making a list of the legaltech companies that have openly announced they are working with LLMs, and a week ago on Twitter I asked for public input on that list. I was interested not just in which companies are using it, but also for what. This has allowed us to put together the most complete list we’ve yet seen of companies using GPT or LLMs for legal use cases, and sheds light on what those use cases are in practice.
The primary use cases we are seeing for this technology in legaltech right now are:
- Drafting (of clauses or first drafts of documents), and
- Legal research.
There also several products that are being developed to support lawyers in multiple ways, as a kind of “virtual assistant”.
I will set out the list below, but first, a few qualifiers:
- Although we at Legaltech Hub are profoundly excited by this development in the technology space at large and the potential impact of it in the legal industry, we do not believe that all legaltech or tech companies should be using GPT or LLMs, nor are we making any kind of qualitative assessment about those companies that are or are not using this technology.
- We are tracking this data because it is of interest to many people, and it will be telling over the next year to see what kind of success the companies that started early to use this technology enjoy. We think the relative success (or not) of early adopter companies will provide an indication of the likely impact this technology will have overall in our industry.
- The list I’ve gathered is based on information that is publicly available, and predominantly highlights companies that have already made announcements that they are using LLMs for particular use cases. We are aware of several other companies that are experimenting with this technology but have not yet made public announcements. We are also aware of several law firms that are experimenting with this technology, again without publicly announcing it. In other words, this list represents only the tip of the iceberg and we imagine there is far more activity happening under the surface that most of us can’t yet see.
- We have added a new feature on Legaltech Hub that allows users to filter solutions by whether they use GPT or other LLMs. Simply click on the search arrow leaving the search box empty to see all solutions, and then apply the “GPT / LLM” value under the Attributes filter. We will continue to track this space, and will be updating the list below and the data on our site.
Without further waffling, here is the list I’ve compiled of legaltech companies (or law firms) using LLMs, categorized by use case type.
|Company / Product||Use Case||Link to Source|
|Summize||Using ChatGPT to auto-generate contract summaries||https://bit.ly/3KiLAsL|
|Docket Alarm||Using GPT-3.5 to auto-summarize PDF litigation filings in dockets||https://bit.ly/3Z612wx|
|Predictice||This French legal research product uses ChatGPT to generate automatic summaries of court decisions.||https://bit.ly/3Y6whXk|
|DocDraft||Uses GPT to turn client notes and previous cases into first drafts in minutes||https://www.docdraft.ai/|
|Spellbook by Rally||Uses GPT-3 to review and suggest language for your documents or contracts in Microsoft Word||https://www.spellbook.legal/|
|Henchman||Uses GPT-3.5 to enrich drafting options, for example to change single to plural in a clause or add an element to a clause.||https://bit.ly/3Zx0r7I|
|Ironclad||Worked with OpenAI to create an automated redline feature, with GPT-3 automatically generating clause suggestions and redlines in contracts that users can accept or reject with one click||https://bit.ly/3EiR9mS|
|Lexion Contract Assist||Uses GPT-3 to help lawyers draft, negotiate, and summarize contract terms. Contract Assist auto-generates clause language, inserts clauses from a playbook, produces suggested redlines, and summarizes clause language.||https://bit.ly/3XBspgI|
|Contract Works by Onit||Contract Works has developed two new features, Clause Creator and Simplify, using GPT-3. Clause Creator auto-generates a clause during the redlining phase based on user specifications. Simplify takes any clause and reproduces it in simpler language to reduce “legalese” and complexity.||https://bit.ly/3k4QwXw|
|AxDraft by Onit||Interestingly, AxDraft, another Onit company, has also created two features using GPT – one called Clause Creator and one called Simplify. They do the same thing as the equivalent features in ContractWorks.||https://bit.ly/3KfWMGt|
|Arteria||Canadian CLM company Arteria is using GPT technology in parts of its end-to-end contracting solution.||https://www.arteria.ai/|
|Malbek||Uses GPT-3 to rewrite clauses in plain English to make contracts more accessible to business users in the contract lifecycle||https://www.malbek.io/|
|Jurisage MyJr||MrJr is a JV between Jurisage and AltaML using GPT-3.5 to allow users to ask a legal research question in plain language and get a quick, plain language answer back that synthesizes case law.||https://bit.ly/3XFZ2tM|
|Lexata||Uses GPT 3.5 to provide clear, accurate answers to complex securities law questions. Lexata draws on a curated database of securities laws to generate answers and surfaces the relevant sections of securities law to users alongside the answer.||https://bit.ly/3IbZWZh|
|Blue J Legal||Blue J is developing a new research product called “Ask Blue J” that will launch soon and uses ChatGPT across curated, current tax and legal domain content, cross-referencing the answers with legitimate sources, to produce a bespoke research memo on a user’s specific query.||Sign-up for early access: http://www.askbluej.com/|
|Alexsei||Using GPT-3.5 (in combination with other models) to respond to research queries by aggregating and summarizing relevant sources and auto-generating a memo.||https://bit.ly/3YYCjtO|
|ScotusAI by Standd||Uses GPT-3 and a curated database of SCOTUS opinions to provide a chatbot that users can ask any legal question of in order to find out what the Supreme Court has said about it.||https://standd.io/scotusai|
Classification / Tagging
|Fastcase, Docket Alarm||Using LLM to programmatically extract classification tags for use cases and then populating a database so that the tags can be applied for use in downstream tasks.||Status: experimenting|
|SALI Alliance||Using LLM to programmatically extract classification tags for use cases and then populating a database so that the tags can be applied for use in downstream tasks.||Status: experimenting|
Search / Knowledge Management
|Standd||A new start-up using GPT-3 to search across lawyers’ own internal work product in response to a query, helping lawyers find the most relevant paragraphs in their previous work and providing links to the documents they’re part of as well as contextual information about the search.||https://standd.io/|
Virtual Assistant / Multiple Use Cases
|Casetext CoCounsel||An AI legal assistant based on the most advanced OpenAI LLM to which lawyers can delegate work including legal research, reviewing documents, preparing for a deposition, reviewing and analyzing contracts, extracting data from contracts, reviewing documents for compliance with a company’s policies, searching through databases, and retrieving previous work product.||https://bit.ly/3y87B68|
|LawDroid Co-Pilot||Uses GPT-3 to provide a virtual assistant that can research legal issues, help draft emails and letters, summarize documents, translate, or “just have a chat”||https://bit.ly/3YYARrm|
|Harvey||Harvey, the tool rolled out by Allen & Overy, apparently uses GPT-4 to help lawyers automate contract analysis, due diligence, conduct research, and generate insights, recommendations, and predictions across multiple practice areas.||https://bit.ly/3k6w2xv|
Companies that have divulged they are “Experimenting with LLMs”
|Syntheia||“experience using GPT and building with it, but not sure it will form part of standard products anytime soon”||https://www.syntheia.io/|
|GenieAI||“have been using GPT-3 and LLM for a year or so now, some very fun features coming soon”||https://www.genieai.co/|
|Mindpeer BD||“we've been conducting experiments and are planning on adding GPT powered features in the next release. We are going to use GPT to summarize material and to supplement reports.” Mindpeer is also working with a different LLM.||https://mindpeer.com/|