Over the past 18 months, Purpose Legal’s advisory team has consulted, perhaps more than any other e-discovery topic, on AI-powered document review. Those conversations move very quickly to the same question: first and foremost, case teams want to know how this evolution of legal AI can help them deliver client value.
Generative AI-based issues analysis, which is one use case for Relativity aiR for Review, has proven particularly capable of doing this. We’ve found that it reduces review time on average projects by 60-80 percent and saves tens of thousands of dollars, while delivering superior results. Larger reviews bring larger savings.
Additionally, issues analysis is very flexible—so it brings cost savings and improved results to a wide variety of real-world e-discovery and investigative scenarios. Allow me to share some specifics.
Going Beyond Relevance Review
aiR for Review’s relevance analysis provides a simplicity that makes it the right fit for many cases. But when you need more, issues analysis may be the answer. A carefully designed issues analysis workflow can provide relevance classifications plus additional analysis that may be helpful or even required for your review.
We have a growing list of use cases where an aiR for Review’s issues analysis consistently improves results, drives efficiency, and reduces costs. For example:
- Identification of documents relevant to multiple key or hot issues
- Grouping specific documents for further review before production
- Removal (from production) of particularly nuanced, non-relevant documents
- Request for production correlation
- Focus for human-in-the-loop prompt calibration
Here’s one illustrative example. Recently, a client inherited a matter from another firm, with a pre-established production deadline. The court rejected attempts to move that deadline and we found ourselves facing a large document review, with ongoing data collections (more review), and the need to complete it all within one week.
All of this made aiR for Review the way to address a complex project and one that would be incredibly difficult to meet in the needed time frame using manual approaches. With production requirements that included document-by-document RFP correlation, the tool’s issues analysis was a fitting choice.
Purpose Legal’s team worked with a single subject matter expert (SME) attorney on the case to develop and execute a workflow that:
- Successfully developed analysis input (prompt criteria) in 3 iterations in just 48 hours.
- Reviewed 300,000+ documents from rolling collections.
- Completed the production on time (with RFP correlation).
- Validated review accuracy with an estimated recall above 95 percent.
In addition to meeting our deadlines, this process reduced review time by more than 80 percent and saved an estimated $70,000 over a traditional technology-assisted review (TAR) workflow. Learn more about it here.
Digging Deeper into the Production Set
We regularly find the need to perform additional focused review within a responsive production set. In this context, aiR for Review’s issues analysis can be a valuable addition to any type of first-pass review (linear, traditional TAR, or AI-powered).
In another project we worked on, the case team had used a TAR workflow to identify a document production set. Their original intent had been to conduct a second-pass review, correlating RFPs with the production documents (a production requirement). We suggested issues analysis as an alternative to attorney review for this second pass.
In five days, Purpose Legal’s advisory team and a senior associate SME used aiR for Review’s issues analysis to complete that RFP correlation for a production set of more than 70,000 documents. These results were faster, more consistent, and less costly than a fully attorney-performed review would have been.
In another recent example, the case team was particularly sensitive to over production of documents related to one specific issue. aiR for Review’s issues analysis allowed the case team to rely on AI classification for most of the corpus, reliably focus SME attorneys on a second-pass confirmation review for the particularly sensitive issue, and use statistical validation for the full process.
Making the Most Out of Your Issues Review
There are a few key considerations to be aware of when using issues analysis versus aiR for Review’s other analysis types, like relevance or key document identification.
First, the analysis inputs leverage up to 10 distinct issues. While it is common for reviews to involve more than 10 issues or requests for production, we’ve yet to encounter a review where we could not, in collaboration with the case team, consolidate to 10 (or fewer) Issue or RFP groups. The effort to accomplish this has generally been minimal and well worth the value returned in refining the focus areas to ensure optimal output.
Second, crafting issue-by-issue analysis input can be a little more time consuming than general relevance input. But there are some general guidelines that lead to quicker development and more effective input.
Finally, working with up to 10 distinct analysis results adds a little complexity to later accommodate in downstream processes (document production, pre-production QC, review validation, et cetera). This is all very doable and, in many cases, very much worth it.
Case Team Reactions and Confidence
Adoption of generative AI in e-discovery document review may be in relatively early stages, but it is taking off and moving fast. Why?
Because it works—and because it is transparent.
Unlike traditional TAR, aiR for Review shows its rationale, cited sources, and considerations, and explains classifications in plain language, making it easy for case teams to trust and refine the results. Case teams can readily see what the AI is “thinking” and interact with it in ways not too different from the way they would with human reviewers.
After seeing the results firsthand, case teams express strong confidence in using aiR for Review for issues review. They consistently describe it as a valuable “force multiplier” allowing them to deliver more value to their clients, in less time, and at reduced expense.
Having witnessed these capabilities in action, case teams see aiR for Review’s flexibility and actively seek new opportunities to leverage its capabilities, capturing the same measurable benefits on new matters and in new ways.
Graphics for this article were created by Caroline Patterson.
