iManage’s Head of Data Science Jan Van Hoecke explains below what generative AI is, and what it can do for the legal industry.
The use of tools such as chatGPT is causing a stir. These tools are able to produce reams and reams sophisticated text by simply asking a few questions. What does this exciting development mean for the legal sector? What are the possible applications of generative AI in law firms and corporate departments?
AI Different from Other AI
It is important to know a little about generative AI, and the large language model that supports it.
The large language model, unlike machine learning models that are trained to perform certain tasks, and have been used in the legal sector for many years to perform tasks such as document classification, consumes enough content to create a “worldview” of sorts, from which it can then draw to generate new content using what it has already read and absorbed.
How can you be sure that the large language model is accurate?
A process called “grounding” ensures that the answers or content generated by generative AI is based on quality material, such as what is found in a firm’s precedents or document management system. This process of ‘grounding’ prevents generative AI going on wild flights when it is put into service.
Unlock Knowledge Effortlessly
Access to knowledge is one of the first areas in which generative AI could potentially be valuable.
What if, for example, a knowledge-management system (KMS), had a ChatGPT interface that allowed users to ask questions and the AI provided an answer based on the most relevant documents from the repository or external legal knowledge sources. It would be helpful if a lawyer was asked to list all the rules regarding an employee’s rights to work in California. Instead of scouring through the many documents within the system, a lawyer can ask the chatbot directly the question to get an accurate answer. The answer will include the source used in forming the answer.
The use of generative AI has been gaining popularity. Tools like ChatGPT have amazed users with their uncanny ability, using only a few simple inputs from the user, to produce reams and reams sophisticated text.
Can its answers be trusted, even though generative AI is willing to take on this type of task?
We return to the grounding concept and emphasize the importance of feeding the large language model with quality input to generate answers and content. The grounding feature ensures that, instead of asking the language model to give its opinion based on its “worldview”, it will actually provide an answer that is based upon the text in the documents the tool was pointed towards. This gives users greater control over AI results and more confidence in their outputs.
Less Difficulty Around Drafting
What if generative artificial intelligence could not only help with finding and drafting legal agreements, but also search for legal knowledge? This is yet another possible use for generative AI.
It is important to ensure that generative AI uses the standards the organization considers the best for writing. Here, firms should really rely on their knowledge and use it. Does the firm have a template of a share-purchase agreement for midsize tech startups, which has been approved by subject matter experts? Consider this as the gold-standard that generative AI can use to help draft that type of share purchase agreements.
Does this AI assistance mean that humans are no longer needed to draft legal documents? It just means more grunt work will be taken out of writing.
The lawyer can ask the AI chatbot, while drafting a contract, to compare the clause in question against the current “market standard” as determined by experts within the organization. In some cases, the lawyer could start with bullet points in a legal contract and have the generative AI access the right internal or external resources to help flesh out the clauses. The human still formulates the intellectual foundation of the legal agreement, they just get some help with the wording.
It is important to ensure that the AI utilises the standards the organisation believes to be best when it comes writing.
Enhanced eDiscovery
The ‘new kid’ on the block, generative AI, has a lot of potential in eDiscovery, as it is one of the most AI-driven processes within the legal field.
Traditionally, eDiscovery relies on supervised machine-learning models to automatically tag and classify data as it combs through reams upon reams of possible evidence. Machine learning models are trained by humans who show them what types of evidence belong in which bucket. Eventually, the model is able to tag and classify with high accuracy.
In this case, generative AI can help in a number of ways. Imagine a situation where a document has been found and automatically tagged in a particular way during eDiscovery, but it is unclear whether the file will be relevant or useful to the case. A legal professional can use a ChatGPT interface to ask AI if the file is relevant or useful. The AI will then decide whether or not the file should be kept.
Generative AI can also help to streamline the eDiscovery by automatically creating summaries of documents such as contracts, legal briefs or deposition transcripts. This allows busy legal professionals to quickly and efficiently understand the important information in large amounts of text.
Operational Efficiencies
It is possible to use generative AI for other smaller tasks in a legal organization, particularly the workflows around matter management. Consider how much time you spend on client reports or summarizing what was said during a meeting or conference call. It is basically gruntwork, but it’s a great candidate for generative artificial intelligence.
Legal professionals can also use AI to help them draft legal documents. They can give AI key points or notes and AI will turn these into complete paragraphs. What is the result? The result? Lawyers are able to spend more time doing the things they love, rather than boring tasks such as reporting or summarising meeting minutes.
What if, for example, users could ask an AI interface within the DMS of the firm a question about e-billing or data handling concerns for a specific client? The AI bot can provide the answer without the need for the lawyer to spend hours poring through the Outside Counsel Guideline file. This will speed up the internal process.
The generative AI can help law firms on both the front-office and back-office sides of the business. It will make it easier to interact with the IT helpdesk, which in turn, will provide faster answers to product or IT queries. Together, these front-office and rear-office operational efficiency improvements can have a significant impact on an organisation.
From Scepticism to Achievement
The legal community may be hesitant to accept generative AI and wonder if is even necessary. It would be a grave mistake to dismiss generative AI as a technology that has no practical applications for lawyers.
Although generative AI is still in its early stages, use cases for it are already emerging, especially in the legal field. Principles like grounding help ensure that its outputs will be accurate and reliable. It would be a good idea for law firms and corporate legal departments to look at ways to put this new technology to work within their organisations to make them more efficient, safer and smarter.
Jan Van Hoecke is the Head of Data Science
1 Phipp Street London EC2A4PS, UK
Tel: +44 02038 796080
Jan Van Hoeckeis an experienced computer scientist who is passionate about technology and solving problems. He is interested in the growing amount of data that is available within organizations, and is determined to create solutions that will help these organisations explore, discover, and learn as much as they can from this information.
iManageis an intelligent, cloud-enabled, and secure knowledge work platform that enables organisations to discover and activate knowledge contained in their business communications and content. The artificial intelligence of iManage and its powerful email and document management create connections between data, systems, and people, while leveraging the context within organisational content for deep insights, informed decisions, and collaboration.