eDiscovery software is software that collects the emails, chats and files that a company may have to hand over in a lawsuit, removes the duplicates, lets lawyers review each item, and then produces the relevant ones for the other side. It is also written e-discovery. Discovery is the stage of a lawsuit in which each side can demand to see the other side's evidence. When the evidence is electronic, the stage is called e-discovery.
Discovery is far broader in courts in the United States than in India or most other countries. A company that is sued in the United States may have to hand over every email, chat message and document that touches the case. So most eDiscovery software is sold in the United States, to law firms and to companies' legal departments.
The work has five steps. First, the company preserves the data: it stops relevant data from being deleted, through a legal hold. Second, it collects the data, by copying it from email servers, chat apps, laptops and cloud storage. Archived email often comes from communications archiving, and copies of laptops from digital forensics tools. Third, the software processes the data. It removes exact duplicates, pulls the text out of every file so that it can be searched, and reads scanned pages with optical character recognition. Fourth, lawyers review the documents. Fifth, the relevant documents are produced, which means handed over in an agreed format, each with its own page number.
Consider a maker of medical devices in the United States, sued by patients who say that its hip implant failed. The court orders it to hand over documents from 60 employees across eight years. The collection comes to 3.2 million emails and files. Processing removes 1.1 million duplicates. Search terms that both sides agree on, such as the implant's name, narrow the rest to 6 lakh documents.
Review is the expensive step. In review, lawyers decide which documents are relevant to the case and which are privileged. Privileged documents, such as advice from the company's own lawyers, do not have to be handed over. Traditionally, lawyers read documents one by one and were paid by the hour, so review was among the largest costs of a lawsuit in the United States. Much of this reading has been done by legal-services teams in India, a business called legal process outsourcing.
Today, technology-assisted review (TAR) ranks the documents first, so that lawyers read the most likely ones and can stop early. In the hip implant case, the review finds 85,000 relevant documents. 4,000 of them are privileged, and they are listed in a log instead of being handed over. The other 81,000 are produced. Newer AI models also write a short reason for each decision, and a lawyer checks a sample of them. This arrangement is called human in the loop.
The legal department follows the cost of each review in its enterprise legal management (ELM) system.