Leverage AI to improve review speed without sacrificing quality

In civil lawsuits in the United States, electronic data and documents that can serve as evidence must be submitted by the due date in accordance with appropriate procedures in order for the defendant and the plaintiff to disclose evidence to each other and sort out the issues before the trial. must beIn particular, the process of handling electronic data is called e-discovery, and if a Japanese company does business in the United States, all electronic data held at its headquarters and data centers in Japan will be subject to disclosure as evidence. there is.Once involved in a lawsuit, in order not to create a disadvantageous situation in the lawsuit, it is necessary to quickly and appropriately sort out the necessary information from the vast amount of data.

In light of this situation, KIBIT Automator is an AI-equipped tool that was independently developed based on opinions and requests from various investigative agencies, as well as knowledge and know-how accumulated through FRONTEO's extensive experience in forensic investigations. Centered around the AI ​​engine "KIBIT," proprietary functions enable efficient and exhaustive evidence searches.Among the work called e-discovery, the document review process is the most costly process, accounting for about 70% of the total discovery cost.The challenge for companies in discovery is how to streamline this review process and reduce costs.

In the past, customers and law firms raised concerns about document reviews, such as cost, time, volume, review speed and skills of reviewers.In addition, due to our previous fee model based on man-months, some customers had to compromise on the quality of difficult reviews in order to keep costs down. By implementing the functions introduced below, "KIBIT Automator" shortens the time required for review, realizes document review with quality that is not affected by reviewer's skill or fatigue due to long work, and sacrifices quality. It is possible to improve the efficiency of review work without having to do it.

Product features

Average number of documents that can be checked per hour from 1 to 40

Document reviews are usually performed by lawyers or people working under the direction of lawyers, who go through each document and decide that it is "relevant".It is time-consuming and costly to separate them into "not relevant". "KIBIT Automator" makes AI learn the contrast between "relevant documents" and "irrelevant documents", judges "relevant / not relevant" with stable quality faster than humans, and Reduces review time and costs.

By utilizing KIBIT Automator for document review,
Cut documents that should be reviewed by humans by about 40%

Reduce the electronic data to be reviewed to 1% to 5% by keyword search, and use KIBIT Automator to identify "documents that should be reviewed by human eyes" (about 40%) and "do not need to be reviewed by human eyes (only AI (reviewed in)” (approximately 60%).Significantly reduce review costs and time. (our average performance)

Main functions

Assisted Learning: Using the cut-off support function to judge "documents that do not need to be read",
Significant reduction in human-readable documents

Document reviews are usually performed by lawyers or people working under the direction of lawyers, who go through each document and decide that it is "relevant".With the Assisted Learning function, it is possible to simulate what percentage of the documents collected as research targets need to be reviewed in order to find a certain percentage of "relevant" documents.As a result, the attorney in charge of the case and the person in charge of the company can exclude documents collected as investigation targets as "documents that do not need to be read", or prioritize and investigate documents that should be read. , the survey period can be shortened.

Highlight Sentence: Highlight the part that needs confirmation

When the task is handed over to the reviewer, the parts judged by the AI ​​to be particularly necessary for confirmation will be highlighted and handed over, soIt is possible to reduce the amount of text that reviewers have to actually read.