Use this button to switch between dark and light mode.

How AI is Helping Small Law Firms Level the Discovery Playing Field

January 09, 2019

Only a few years ago, small law firms had legitimate cause for concern when faced with the prospect of having to review thousands—or even millions—of documents during the discovery phase of a lawsuit. How could a law firm of only a handful of attorneys and staff effectively review such a large number of documents that even an army of “Biglaw” associates would struggle with? For many small firms, the answer would have been to hire contract attorneys on a temporary basis to tackle these documents—an expensive solution.

But today, thanks to Artificial Intelligence (AI), small law firm lawyers and staff have a new tool that can level the discovery playing field and help them review massive amounts of documents in a fraction of the time that it used to take—without bringing aboard contract attorneys.


Technology Assisted Review (TAR), which is also known as predictive coding, is a form of AI that can quickly review and code documents based on what the software learns from previously coded documents. While it could theoretically be used in any instance where coding of documents is necessary, such as for determining whether certain documents are privileged, TAR is typically used to determine the relevance or responsiveness of those documents. As most lawyers who have ever reviewed documents know, it is this process of finding needles (relevant or responsive documents) in the haystack (the entire set of
produced documents) that is most time and resource intensive for lawyers, especially those at small law firms.


And that’s where TAR can level the discovery playing field. It can do the work of human document reviewers in a fraction of the time, with outstanding accuracy. In other words, TAR functions as a superhuman document reviewer and coder. Here is how TAR works:

First, lawyers identify a representative cross-section of documents from the universe of documents that is to be reviewed. This subset of documents is often referred to as the “seed set.” The larger the seed set, the more accurate TAR will be with its coding. From there, lawyers review the seed set electronically and code each document accordingly. Next, they input those coded documents into the predictive coding software. The software analyzes the seed set and creates an algorithm specifically for the purpose of that particular aspect of the document review, such as relevance. The results of the algorithm are applied to additional sample documents. Reviewers can help the algorithm “learn” and thus perform better by reviewing these additional documents, inputting the results into the software, applying the updated algorithm to another sample set and continuing the process until the documents are coded appropriately.

Once the algorithm has had this initial “training” and reviewers are satisfied with the software’s performance, the algorithm can be applied to the entire set of remaining documents—whether those number in the thousands or millions. At this point, TAR assumes its superhuman document reviewer role and can do in minutes what an army of document reviewers would need days to accomplish.


With TAR, there are two methods of learning: active and passive. Passive learning is the method described above, which relies on the use of seed sets culled by reviewers to train a software’s algorithm. Active learning, on the other hand, is when the software’s algorithm itself chooses additional sample documents based on what it believes are relevant, responsive, etc., which are then manually coded by a reviewer. When active learning is used, the importance of seed sets decreases because the software’s algorithm is always in a state of learning.


With the advent of any new legal technology, there is often hesitation on the part of lawyers to dive in without state ethics regulators or courts first giving their blessing to the new development. When it comes to TAR, courts have looked favorably on the technology going back to early 2012 when then-Magistrate Judge Andrew Peck issued his decision in Da Silva Moore v. Publicis Groupe & MSL Group, 287 F.R.D. 182 (S.D.N.Y. 2012). Judge Peck’s Da Silva Moore decision marked the first time that a judge had accepted the use of TAR.


The use of AI in discovery—in this instance, TAR—has game-changing implications for small law firms. No longer must small firms be concerned about how to handle an enormous upcoming document review or the impact it would have on their overall practice management. This means small firms no longer need to worry about finding a team of contract attorneys (or the resources to pay for such a team) to tackle a review. And, small law firms need not be worried about being “outgunned” by larger opposing counsel who have teams of associates to assemble and produce large volumes of documents.

With TAR, small law firms can be on equal footing with their larger adversaries during the discovery phase and can be better advocates for their clients because of it.