iManage today (21 August) unveils iManage AI, a new AI engine built natively into its knowledge work cloud platform leveraging Microsoft Azure OpenAI Service.
The core of the new iManage AI solution is a document classification and enrichment engine, which automatically analyses documents and extracts key data points, such as jurisdiction, parties, or dates, saving that information with the document. It is also capable of automatically tagging emails and filing them in the right place, and it enables users to have ‘a conversation’ with their data using a ChatGPT-style interface.
Speaking to Legal IT Insider, iManage’s VP of product management, Shawn Misquitta said: “iManage AI is a set of services designed to help the knowledge professional either become more efficient by taking away their mundane tasks, or become more knowledgeable or smarter by asking questions of the data that traditionally would have taken them reading the entire document or searching for different phrases in the document.”
Two customers are using the classification engine in production, and Misquitta said: “We have several others that are enabling it as pilots. Alex [Smith] and his team have spent thousands and thousands of hours working on highly curated datasets to train this AI on. And that’s really where we are different in terms of the accuracy of the classification, which is over 90%.” The engine can detect around 30 different contract types and that will be continually expanded.
An Trotter, senior director of operations within the office of general counsel at Hearst, said: “For a long time, our attorneys have reported difficulties with search, primarily to retrieve language that fit a particular use case, and also to find specific documents. So, we have used iManage AI to apply metadata values consistently to assist with retrieval and search. iManage AI has automated the process whereby documents in iManage Work are analyzed by the AI engine, which makes an accurate prediction on the type that each document is, applies metadata and pushes them back into the document management system so that they are searchable.”
With regard to the new email functionality, Misquitta said: “The rate at which customer data is growing, it’s practically doubling every five years or so, and email is the majority of it. With e-mail you have attachments. So there’s a lot of knowledge also in e-mail but knowledge workers don’t go to law school to sit down and file emails for one to two hours a day. So we said, can we completely take this away from you by looking at the collective of communications that exists inside iManage in the knowledge file and then we can look at your mailbox and automatically, just like Microsoft Viva does in the morning, give you a list and say ‘this is what I think you should do. These are the things you need to address. So I’ve taken care of these things and filed these emails for you.’”
Misquitta added: “It learns from organisational behaviour, but more importantly it uses the iManage context of clients and matters to help the lawyer wade through the ton of e-mails they have in their mailbox.”
With regard to the third use case, iManage is leveraging generative AI to enable users to ask questions of their documents using natural language. Misquitta said: “So instead of having to read through the documents you could say ‘tell me what are my obligations are for this’ and it will extract it and create a report for you.” Sources for the answers are provided alongside and Misquitta said: “Citing references for the recommendation is incredibly important to the industry, so they are not just taking it at face value.”
The data science team has worked on reducing hallucinations so that if the answer isn’t available, it will say so. Additionally, the AI engine only analyzes documents located in the customer’s own data resource to ensure confidentiality and compliance with ethical and regulatory obligations.
To find out more, see iManage at ILTACON, booth 615.
So, AI will sort the emails and make a suggest to file them in a group, then there will be that one email that should not be filed into the DMS, that will slip through. It will be filed because a person did not directly review it and make a decision to file it. Then, that errant email will be provided on discovery into the wrong hands. And that will be the end of using AI to help file emails. Humans need to handle emails and documents and file them. AI is not going to recognize an email that has negative comments or internal discussions about the client. Attorneys will trust AI too much and after a while, that will be the habit. The AI focus should be on eliminating the amount of emails, imho.