In our last blog posting, we discussed the fact that predictive review can reduce costs, improve quality and sometimes even save your case but many legal professionals don’t use it because they don’t trust it. A second reason is that they believe predictive review is too difficult and too much of a burden on attorneys at the start of the process
The traditional predictive review process begins with attorneys reviewing enough documents to create a seed set. (These are the human decisions that the tool applies as we discussed in part one.) By its nature, this work must be done by at least mid-level attorneys. With the billing rate for an associate level litigator now ranging from $288 to $403, according to BTI Consulting, many believe that their time is too valuable for such work.
Other attorneys may have other objections.
Some feel that reviewing documents for responsiveness is beneath them, and some are concerned they do not have enough information about the document set or familiarity with the document sources to do a proper job.
While it’s true that many attorneys have their reasons for not wanting to put in the time to jump-start a predictive review, it is no longer true that they have to. iCONECT’s predictive review tool does not require attorneys to seed documents the traditional way at the beginning of the process. Instead, users have the option to start review with patent-pending Xmplar® which enables attorneys to take language from the complaint to create a customized “perfect document” as a search model or “exemplar” for the predictive review tool. The attorneys will also receive “SmartSamples” to start the seeding process. SmartSample is a revolutionary way of gathering batches that are representative of the entire database. It ensures that all the right concepts are included in the first few document that are reviewed. All in all, predictive review can now commence with almost no upfront time investment—and no time spent reviewing documents to create a seed set—from attorneys and it can end faster by eliminating so many false positives and cutting out the number of iterations needed to train the system thoroughly.