Counsel, Head of Legal
Text IQ has applied artificial intelligence to privilege review, a process that is historically risky, inefficient, and costly. The diagram below demonstrates the literal transmutation of privilege review with Text IQ.
On the left, the traditional process begins at the outer circle. Teams apply search terms (or a “privilege screen”) that divide the document population into two parts: potentially privileged (PP) and potentially non-privileged (PNP). Human reviewers review the PP population and find non-privileged (NP) documents that they code as ready for production. PP documents not marked NP are sent to second level privilege review and logging.
Manual reviewers also sample and review the PNP population, and invariably they find privileged documents, in part from finding new attorneys and new law firms that weren’t on the initial privilege screen. This new information generates new search terms, which are now fired on a smaller document population. The process repeats itself, as shown in circles two, three, and beyond—in fact, reviewers can never be certain that they have found every privileged document until they run this sequence a very large number of times.
Compare this to the process on the right, where Text IQ analyzes every document and surfaces unknown privilege terms, attorneys, and law firms before the population splits, preventing the spiral from happening. Two populations result: an NP population that is so accurate it can be produced after you apply your existing, defensible QC process; and a PP population that is weighted, prioritized, and explained to power an efficient second level review.
Scores and Reasons
Text IQ builds interpretable AI. We have built a human-readable “semantic layer” into our unsupervised machine learning models, powering two features that are unique in the marketplace: Scores and Reasons. Scores are automatically generated for potentially privileged documents, ranking the machine’s predictive confidence and powering a prioritized second level review by outside counsel. Reasons are also generated in the form of sentences in natural language that explain why documents have been flagged as privileged, which simultaneously automates a first draft of the privilege log.
The protocol for document review includes identifying both privilege and responsive information, and Text IQ for Legal is a purpose-built solution that can automate and augment both. In addition to our solution for privilege, Text IQ builds a solution that automatically generates Scores for responsive documents, validated by statistical modeling. Text IQ for Legal also discovers documents that can create a reputational disaster. If it’s memorialized in writing, we will find it.