Q&A: Using Generative AI to Train and Refine Technology-Assisted Review Workflows

In addition to IQ.AI by FTI Technology, our teams use a variety of generative artificial intelligence tools to improve outcomes across all phases of clients’ investigations, compliance exercises and e-discovery matters. As one of our technology collaborators, Reveal provides numerous products that support these objectives. Will you talk a bit about how we’re working with Reveal’s platform?

The power of Reveal is within its ability to provide attorneys and corporations with immediate access to insights across the spectrum of their data. With its retrieval augmented generation and other generative AI capabilities, we can provide context for the prompts that help classify documents as responsive, not responsive, hot or not, privileged or not. Drawing on these classifiers, we can calibrate through sampling, pointing the tool at a pool of documents that we know contain relevant material and using the feedback to further hone the model and improve the classifiers. 

From there, we can use the refined classifiers to expedite training of a traditional technology-assisted review or supervised learning model. This allows us to train a model for a large document review much faster than the traditional approach of a team of lawyers training it manually. That time can make all the difference in meeting deadlines in time-sensitive matters. It can also contain costs for our clients and allow us to more efficiently conduct quality control and validation steps. 

What’s unique about using RAG within Reveal?

RAG is very beneficial for many different use cases and is available in other generative AI solutions for e-discovery and investigations. What’s unique in Reveal is that it is embedded in the platform by default, so we have access to use it where needed without any additional add-ons required. With the RAG features built in and ready to use, we can easily incorporate it into the other workflows in play, which supports our approach of being highly flexible and adapting solutions to meet any client need. 

Reveal’s recent e-discovery buyer’s report found that 100% of legal technology decision makers said the ability to run their own fine-tuned AI models within their e-discovery environment is important to their operations, so this is clearly a feature that matters to our clients.

What are some of the other AI features within Reveal that stand out?

With Brainspace integration, we can draw on the entity extraction that’s done automatically as part of the initial ingestion into Reveal. This gives us an organized view of specific case elements. Surrounding relevant entities (e.g., topic areas, regions, individuals), we can look at related documents and examine sentiment (was an exchange persuasive or inappropriate, etc.) to help discover instances of red flag behavior. 

In an investigatory context, the ability to see the communications patterns across a set of documents to surface key facts more quickly. For example, we can easily see how frequently certain individuals talk to each other and who else they’re in contact with, to identify how information is flowing between persons of interest. In a Foreign Corrupt Practices Act investigation, where communications patterns can provide important indicators and evidence of misconduct, this capability can help us find that information faster.   

Can you share an example of a matter where these capabilities made a difference?

In one recent engagement, we partnered with a law firm to download all publicly available information from a large U.S. Department of Justice investigation and used Reveal to identify whether specific individuals were named in the investigation. This helped the law firm prepare a response strategy for any of its client companies that employed individuals who were potentially subject to further inquiry. 

Another example is a recent intellectual property theft case. We used a combination of Reveal’s generative AI tools, keyword search and communications analysis to determine the activities of an employee who had departed and was suspected of taking sensitive materials and IP to a new employer. This is a common type of case for our digital forensics investigators, but in this matter, we added the analytics and Reveal’s RAG to quickly piece together fact patterns that confirmed our client’s suspicions and supported the case.

One more example is a matter in the U.K., where we were working with a client following a dawn raid. The client’s law firm wanted to conduct a helpful/harmful review to understand what was in the documents seized by the government. Using Reveal, our team quickly surfaced the key themes and provided counsel with indicators of what avenues of interrogation regulatory agencies might pursue following the raid.

The Reveal report you mentioned also found that 79% of legal technology buyers said that access to the latest AI and analytics is one of the most important factors they consider when deploying tools. To that end, is there anything else you would add about how our expert teams use Reveal?

Generally, these features, and the way we use them in tandem with other technologies, are all about increasing the speed at which we can establish a clear and rich understanding of a data set. They are also complementary to our suite of IQ.AI solutions. Because our approach is technology agnostic and flexible within a number of platforms, we can execute investigations of any size efficiently.

I’ll also add that the report found that security is the number one priority for legal technology buyers, and this is an area we have also made a top priority for our clients. When testing and evaluating a potential technology partner, we engage in a complex and exhaustive review of the company’s technical and organizational measures, performance and contractual agreements in relation to privacy and security. When prospective partners provide AI or large language models, security and privacy review processes are further enhanced to ensure proper use and handling of data within analytic models. So, we’re keenly aware of upholding security and privacy for every engagement and within every technology platform we use on behalf of our clients. 

Related topics:

The views expressed herein are those of the author(s) and not necessarily the views of FTI Consulting, its management, its subsidiaries, its affiliates, or its other professionals.