On Feb. 24, 2012, Judge Andrew Peck issued his opinion in Da Silva Moore v. Publicis Groupe, 287 F.R.D. 182 (S.D.N.Y. 2012), the first case to approve the use of predictive coding in a litigation matter. Many legal commentators declared that this was a death knell for the use of traditional human document review to identify relevant documents in litigation. Predictive coding could replace human intelligence with artificial intelligence, allowing relevant documents to be identified more quickly, less expensively, and perhaps even more accurately than human review. Some companies and law firms (including our own) invested in predictive coding software.

Now, three years later, human document review remains very much alive—indeed, far more prevalent than predictive coding. Given all of the reputed advantages of predictive coding, why is that still the case? This article explores seven barriers that have slowed the adoption of predictive coding as a substitute for human review.

Not Enough Documents

To train predictive coding software first requires human review of a representative sample of the documents. Depending on the case, the documents and the software used, this can require human review of thousands of documents. If only a small volume of unreviewed documents remain after training, then most cost and time advantages of predictive coding have been lost.