At DISCO, AI has been used for document prioritization since 2015. And a 2022 IBA research report states that 40% of young lawyers believe AI and legal technology is critical to the future of their profession. Now generative AI, although relatively new to the game, is taking tech-enabled document review to another level. This article will cover how to use generative AI for document review.
Why use generative AI for document review?
Benefit: Rapid information summarization saves time and money
As the channels we use for communication have grown more technical and complex, so has discovery. The complexity stems not only from the proliferation of various communication methods, such as email and short-form messaging applications like Slack and Microsoft Teams, but also from the sheer volume of information.
In contrast to previous AI-enabled review methods, which involved scores and prioritization, lawyers can treat generative AI just like a junior associate: Write a document review protocol with descriptions for each tag, show your generative AI model each document that needs to be reviewed, and ask it, “Does this document meet this tag description?”
There are two clear benefits to using technology to do this first pass, rather than humans: It can be done at a fraction of the cost, in a fraction of the time. (Generative AI-powered document review platforms can review at speeds of thousands of documents per hour.)
Generative AI can also do things easily that would be a huge lift for attorneys, such as provide narrative justifications for each decision. These justifications can provide attorneys with peace of mind, and — in the case of privilege — can be dropped into a privilege log with minimal editing.
DISCO Vice President of Product Strategy, Katie DeBord, says, “Generative AI is going to allow law firms to get through large document volumes much more efficiently than before, saving their clients time and money. I expect many clients to push their attorneys to adopt tech like this in the years to come.”
Example: Generative AI-powered document review in action
Samika, a junior associate, is working on a case with over 100,000 documents. Rather than tagging documents to provide training examples for an AI model (or, heaven forbid, reviewing documents one by one herself), she can leverage large language models and generative AI.
How? Samika can write the same kind of document review protocol she would write during any other review, and simply give the context and definitions for each tag to a tool powered by generative AI.
For instance, she can tell the tool that “privileged” means:
“Any communication between a lawyer and their client discussing legal matters.”
The tool will then read each document, decide whether the document meets that criterion, and tag the document accordingly. Using generative AI, the tool can also provide a natural language justification, such as:
“This document is an email between Attorney X and Client Y discussing whether they should accept revisions to a contract.”
If Samika is using a tool like DISCO’s Cecilia Auto Review, her entire review can be completed in a week – even if she is the only person on the team.
Generative AI-powered document review: Challenges and solutions
Challenge: What if the AI tool misses something critical?
It’s hard to escape the fear that if you don’t review every single document manually, you’ll miss a piece of information that could be vital to your case.
Solution: Hybrid AI-human review
Generative AI does not work in the same way as TAR 1.0 or 2.0. Unlike TAR – which relies primarily on an iterative cycle of machine suggestions and human review – generative AI relies on well-crafted prompt engineering to quickly categorize documents according to highly specific issue tags.
With DISCO’s Cecilia, attorneys can benefit from an offering called Auto Review. This leverages Cecilia’s ability to instantly consume massive amounts of information, and comprehend natural language instructions to identify whether the documents at issue meet certain definitions.
Auto Review, like TAR, does rely on a layer of human intelligence and confirmation. It begins with a human attorney crafting a prompt – such as, “all communications between Sales and Compliance related to Deal X,” and reviewing the results on a significant number of documents.
If the team decides the documents are perfectly categorized, the rest of the document population can be auto-reviewed. Otherwise, the prompt may be adjusted, and the process will be iterated on until the team is satisfied with the AI’s results. Only then should you run the tool on the entire document set.
It is important to note that the same kinds of quality control (QC) that have traditionally been a part of TAR 2.0 reviews should be applied to Auto Review. By layering in estimation samples, validation samples, and other kinds of statistical QC — either on their own or in partnership with a service like DISCO’s managed review — attorneys can feel confident in the results.
Related reading: Managed Review FAQ
Challenge: Ethics and compliance
Legal professionals are required to adhere to strict ethical standards, which include supervising non-lawyer employees, and retaining all responsibility for the legal work done on behalf of clients. Maintaining data privacy is also a paramount concern.
Solution: Choose an AI provider that prioritizes security and compliance
The good news is, technology-assisted legal workflows have long been approved by courts, notably in Da Silva Moore v. Publicis Groupe, 287 F.R.D. 182 (S.D.N.Y. 2012). Generative AI may be novel, but it follows a long precedent of AI being regarded as an accepted – even integral – element of good legal representation.
Additionally, make sure any AI tool you use for ediscovery is designed to comply with relevant data protection regulations. Hosted on AWS global infrastructure, DISCO’s tech upholds the highest standards for privacy and data security. This level of compliance, including SOC2 Type 2 and ISO 27001 certification, as well as GDPR and HIPAA compliance, ensures that sensitive client data is protected at all times.
In addition, select a partner who is familiar with legal precedent regarding the use of AI, and experienced in managing AI-assisted projects. DISCO Ediscovery is expertly designed to manage, process, and store ediscovery data in accordance with these regulations.
Not a DISCO user? Ask your vendor to verify the regulatory compliance of their platform for data handling, processing, and storage.
Related: Case study: DISCO Review Shines with High Accuracy and Increased Efficiencies 💡
Adopting generative AI for document review: Actionable next steps
- Establish clear protocols for reporting AI usage
Certain legal jurisdictions require disclosure of AI usage and purpose. Establishing clear protocols will help you answer any questions. For example, you may want to specify that certain information must be recorded — such as the specific AI tools used, the purpose of their usage, and the outcomes achieved.
- Focus on data privacy and ethical compliance
Implement strict guidelines and data protection measures in your document review process. Select a vendor who is familiar with applicable data protection regulations, and ensure that your chosen generative AI system is hosted in an environment that safeguards the privacy and security of your clients’ data.
Speed up document review with DISCO’s Cecilia
Ready to use generative AI to streamline your document review process? DISCO is here to help.
DISCO’s category-leading ediscovery platform is easy to use, with instinctual search visualization, robust data management, and a securely built global infrastructure. And now, DISCO’s Cecilia AI takes litigation up a notch with Cecilia Q&A (the AI assistant that can answer any question about your case), Cecilia Timelines (AI-generated timelines that summarize key facts in minutes), and Cecilia Auto Review – plus, more exciting advancements to come.
See how we can transform your practice: Request a demo.
This article comes from our downloadable ebook, Generative AI for Litigation: What You Really Need to Know.
Check out the other articles in this series: