Case Study: 4 Ways We Increased Speed for an HSR Second Request

Case Study: 4 Ways We Increased Speed for an HSR Second Request

The parties to a large acquisition received a second request for information from a government agency. One party hired a top law firm to fulfill the request as quickly as possible, and as the time-pressure built, the firm brought in Text IQ.

48% fewer documents to review meant 2,040 fewer hours

2r Case Study 1

The firm applied search terms to a large-scale dataset and returned 106,640 documents, delivered in two batches. Text IQ determined that 48% of the documents were non privileged (NP), creating a 91% increase in precision. Our AI made predictions of NP to such an extent of accuracy that the population could be rapidly produced, after a defensible QC process. By eliminating the need to review 50,920 documents, we saved the firm 2,040 hours (assuming a contract review rate of 25 docs/hour).

Over subsequent datasets, we achieved a speed gain of 30%

2r Case Study 2

Our platform uses a continuous feedback loop to automatically apply new learnings to subsequent datasets, becoming faster and more accurate over time. This acceleration became evident when the firm provided its data in two batches, and we compressed the timeframe on Batch 2 by 30%. Our AI works by understanding organizational idiosyncrasies. We came to understand the customers’ domains, roles, relationships, parsing patterns, disclaimer language, and so-called “priv-breakers.” Then, we propagated this knowledge forward to the second batch. This organizational fluency is the basis for our platform, The AI Brain, which becomes more effective with each matter at a customer.

Scores powered prioritized reviews and workflows, saving 805 hours

2r Case Study 3-1

Text IQ automatically generates Scores for potentially privileged documents and groups them into three tiers based on the likelihood of containing privileged information: low, borderline, and high. Reviewers spent an average of half the time reviewing the low and high confidence populations, which had a combined volume of 40,240 documents. This automation saved 805 hours in total. Scores also prioritized workflows, by allowing our customer to produce as many NP documents as quickly as possible, beginning with the bottom tier. (Alternatively, a customer with the goal of logging privileged documents as quickly as possible might start at the top.)

We created 80% efficiency gains in the privilege log, saving 550 hours

2r Case Study 4

“This is the most expensive part,” the firm said, describing the privilege log. Navigating this sensitive task—creating privilege log entries that achieved the nuanced balance between waiver on the one hand and unacceptable opacity on the other— was consuming the most skilled attorneys’ time. Text IQ provides Reasons, a feature that is unique in the marketplace, to automatically describe the characteristics that make a document likely to be privileged. Reasons help guide human review and also provide the first draft of a log entry. While attorneys had final say over each log entry, Reasons made them 80% more efficient over the traditional process.


The client met the government’s stringent demands, completing the production and the Privilege Log in a timely manner. What they took away was a new conception of “timely manner.”