There’s a lot we hope to leave behind in 2024: vague RFPs, eleventh-hour document requests, and outdated review methods that require juggling responsiveness designations, privilege calls, and case prep manually.
While it may be wishful thinking to expect all these frustrations to subside, our team has been working hard to ensure Relativity aiR helps you make cumbersome review a thing of the past. With aiR for Review and aiR for Privilege both released in 2024, and aiR for Case Strategy coming right around the corner, there’s no better time than now to double down on your new year’s resolution to focus on more meaningful work.
aiR does it all: helping you review for relevance and issues, identify key material, protect privilege, and develop your case narrative. Plus, working within RelativityOne means you can use these fit-for-purpose solutions in the context of a highly secure, flexible, and scalable platform you know and trust.
What does it look like? Move from one phase of your project to the next via this three-stage workflow:
- Use aiR for Review to identify your relevant documents.
- Use aiR for Privilege on those relevant documents and find privileged material to exclude from production.
- Use aiR for Case Strategy to analyze the key documents found during review, helping you prepare for depositions and trial.
Read on to learn the ins and outs of each product and how to start transforming your processes.
aiR for Review: Find Everything That Counts, Fast
aiR for Review is all about identifying the most important documents in your matter, whether you’re reviewing for relevance, tagging for issues, or isolating key material for an investigation.
The very first step is to identify your total review population. Some documents might still be best suited for manual review, such as image files and those without text. Set those aside, and you’re ready to get started with aiR.
As with any generative AI solution, prompt development is a big part of being successful with aiR for Review. To help with this process, we recommend using a sample data set of 50-100 documents you know are important to the matter to iterate with as you begin your project. This will give you time to fine-tune your prompt and verify your results on a known document population.
Tip: Not super familiar with your data set yet? Don’t fret! You can quickly apply other RelativityOne tools, such as clustering, to identify a diverse set of documents to use in your sample.
Once you have sample data you feel comfortable with, it’s time for the fun part: crafting your prompt criteria. Through an intuitive interface, you’ll tell aiR for Review which documents are relevant to your project, and why. Think of it a little like a review protocol, but in plain language (and feel free to add emphasis—aiR won’t take offense at all-caps instructions or abundant exclamation points!).
Run aiR for Review on your sample set to classify which of those documents are responsive. When the analysis is complete, review the output—not just the designations, but especially aiR’s citations, rationales, and considerations—to identify whether aiR misinterpreted anything.
If you identify conflicts between how you and the AI think about a document, provide aiR with some additional instructions by adjusting the prompt criteria accordingly. aiR for Review makes this prompt iteration process easy by providing those detailed rationales and considerations, so you can see exactly why aiR made each decision and quickly narrow in on where it could use more guidance.
Tip: Treat aiR like a human! Imagine aiR is a very intelligent, but junior, attorney. What would that person need to know to find the right documents? Craft your prompt criteria as a set of instructions to do so.
When you’ve fine-tuned your prompt criteria to minimize conflicts, run aiR against a statistical sample (usually about 400 documents) and calculate precision and recall by manually coding and comparing the results.
If you’re satisfied with the metrics, it’s time to use aiR for Review on the entire review population! If there are validation requirements for production, you can validate and QC the full results by choosing another statistical sample, manually coding it, and calculating its precision and recall.
With that, your analysis is complete—and it’s time to dig into the insights you’ve gathered.
Tip: To ensure accuracy, you can calculate the elusion rate, or percentage of uncoded, low-ranking documents that were found to be relevant within your sample set. The lower the rate, the better!
aiR for Privilege: Protect Sensitive Information
Once you’ve used aiR for Review to find the relevant documents in your project, what’s next? Making sure privileged material stays protected—and that’s where aiR for Privilege comes in. To get things started, you’ll take the documents that aiR for Review marked as responsive, include their families, and use them to set up your aiR for Privilege project.
Through a structured workflow and a combination of various AI technologies, including social network graphs and machine learning, aiR for Privilege does the grunt work and reduces the noise in your data.
- First, it cleans your data, removing spam emails and irrelevant disclaimers that can lead to incorrect interpretations of documents and communications.
- Next, aiR for Privilege builds context around the people and entities in your matter. It identifies and categorizes domains, distinguishing between law firms, clients, and third parties while also recognizing equivalent domains like subsidiaries or alternate email addresses. This ensures communications are correctly associated with the appropriate individual.
- Then, aiR for Privilege determines the lawyers in the data set and designates a privilege status (privilege conferring, privilege breaking, or privilege neutral) for each person, depending on their role in the matter.
Throughout this process, you, or the appropriate subject matter expert, will annotate each decision, reviewing and confirming the AI’s thinking, at every stage. While aiR for Privilege does the heavy lifting, you keep control over its decisions.
Tip: Enable aiR for Privilege’s Client Brain, which stores annotation decisions about disclaimers, email domains, entities, and more across projects from the same client. It’ll make future project setup for that client a breeze and produce more consistent results.
Now it’s time to put generative AI to action! aiR for Privilege takes everything it learned from the setup process to identify the privileged documents. It uses the people involved, their roles, and each document’s content to craft a unique prompt that is sent to an LLM.
Every document in your project is given a privilege prediction alongside natural language reasoning and citations that support the decision. This detailed reasoning provides you with valuable insight that you can QC quickly to make final calls, fast. On top of it all, aiR for Privilege generates a privilege log description so you can simplify and accelerate what is often a painful log creation process.
With aiR for Privilege, you’ll move rapidly through first-pass review and make tough calls with confidence that the right information is being protected.
aiR for Case Strategy: Build Your Story
With document review complete and the crucial information for your case identified, it's time to delve deeper into your data so you can better prepare for depositions and trial. aiR for Case Strategy helps with exactly that. It’s currently available in Advance Access and will be fully released later in 2025.
While aiR for Review surfaces the most important documents for teams to further evaluate, aiR for Case Strategy analyzes those crucial documents to identify key facts, dates, and people.
You’ll start by putting the key documents you identified using aiR for Review into aiR for Case Strategy, which will pull out the most essential information, organize it, and create easy-to-understand timelines, fact chronologies, and outlines. You can also use it to better prepare for depositions by leveraging witness summaries and deposition outlines that provide insights into who witnesses spoke to and when.
Tip: aiR for Case Strategy is most effective on smaller sets of very relevant documents. Avoid including duplicate documents, repetitive email threads, and privileged documents that wouldn’t be used at trial.
But it doesn’t stop there. After depositions are complete, use aiR for Case Strategy to prepare for trial, whether through its deposition transcripts, its trial outlines that organize key facts and issues, or its document summaries that highlight what you need to know in your material.
Furthermore, the solution’s visualizations can be used to easily present information to those who need to see key case points in a quick and digestible format—whether they’re senior members on your case team, other colleagues, or your client.
aiR for Case Strategy’s strength lies in storing information in a clear way that allows you and others to stay apprised of the most important facts and figures to make more strategic case decisions.
Tip: aiR for Case Strategy comes with inline editing and document preview capabilities. Use these to quickly edit or organize your facts without needing to open the document viewer or other subpages.
Take Your Legal Data Review and Make It Float on aiR
It’s time to transform the way you work. With Relativity aiR, the tools you need are right at your fingertips. Embrace this opportunity to find what matters, protect your clients, and build your case narrative, faster than ever. These resources are available—there’s no better time to take advantage and get ahead of the generative AI curve than right now.
“aiR is an open door to pay more attention to the discovery process and the documents that you're analyzing throughout that process. It also allows lawyers who are going to be using those documents for downstream use cases to go ahead and think about those earlier in the process, as they're coding for responsiveness, depo prep, and the documents they need to prove their claims or defenses. And finally, it helps them with privilege questions that can be very complex and very expensive.”
Alison Grounds, Managing Partner, Troutman Pepper Locke eMerge
Interested in learning more about how to revolutionize document review? Join us on our upcoming webinar on Relativity aiR for Privilege.
Graphics for this article were created by Natalie Andrews.
