This blog discusses how technology can be leveraged to streamline the eDiscovery process, specifically focusing on improving speed and accuracy. It emphasises the importance of planning and choosing the right technology for each stage, from Early Case Assessment (ECA) till the completion of a matter. The blog also highlights several helpful tools and how they can be used to maximise productivity, including keyword search, clustering, and AI-powered features like auto-redaction and predictive coding.
eDiscovery is underpinned by the Electronic Discovery Reference Model (EDRM). Whether followed in a linear fashion or not the underlying aspects of identifying, collecting, preserving, reviewing, and producing electronically stored information for legal proceedings or investigations are fundamental. Technology can be applied to help with Early Case Analysis (ECA), data overview, speeding up review or ensuring that the document productions are consistent and accurate.
ECA Technology
There will always be the dilemma of how much to collect. The decision as to whether to collect broadly and narrow the data set after uploading the documents to a database or whether to be more targeted in a collection are necessary considerations in consultation with the client and the litigation service provider. The decision can only be made after paying attention to the risks and consequences and requirements. There is no one size fits all and no right or wrong answer. Decisions need to be justifiable and considerations around proportionality will be key. Cabo Concepts v MGA Entertain is a warning to collect carefully and appropriately. In this case about 3 weeks before trial MGA disclosed missing around 84,000 documents with inevitable cost consequences. The missing documents were due to a collection issue.
The technology that is available to assist ECA will help create efficient workflows. Deciding what to review and what not to review can be helped along with various analytics examples are key word searches often better approached using search term reports which provide a helpful analysis of the results of keywords. Clustering provides a high level overview of conceptually similar documents and can help inform what to keep in and what is likely irrelevant. A document type analysis helps identify how to best review documents, which may require manual review and what issues can be tackled upfront.
At this stage understanding document types and numbers, resourcing and timelines will help identify the most appropriate technology to create efficient review.
Review
ECA will provide many data insights and that will help inform the workflows.
If emails are a dominant part of the data, consideration should be given to email threading and how that can be utilised. Email threading allows for efficient review by ensuring that conversations are collated together. It allows whole conversations to be assigned to specific reviewers creating efficiencies of speed and consistency. With email threading it is possible to determine that just inclusive emails can be reviewed, that is just the emails with unique content.
If there are multiple languages in the document set, using a feature like Language Detect can provide the breakdown of which languages are in the document set. Using AI translation tools documents can either be translated on mass or singly into a language that the reviewers can understand and the document retains all of its original formatting. This is often more efficient than trying to put together a review team that has multi lingual skills.
Redaction is critically important to protect privacy and to maintain confidentiality. It can often be time consuming and fraught with issues of consistency. Tools such as auto redact which find words, phrases or paragraphs, or specific PI such as credit card numbers or tax file numbers can be used for speed and accuracy. Auto redact will find and redact. The words or phrases can be customized.
Categorisation:
Categorisation is a useful feature that can assist categorise conceptually similar documents with the help of a human reviewer providing some good examples. It is useful to be able to organise a document set. The final outcome can be used to help form knowledge teams,
Predictive/Prioritised Review
TAR, CAL, TAR 2.0, predictive coding are all now familiar AI techniques for getting through a document review set efficiently. The predictive coding model uses human decisions to predict how a document is likely to be coded. The most commonly used approach is TAR 2.0 or known as CAL an acronym of Continuous Active Learning where the algorithmic model constantly updates as the review takes place. CAL can be used for both a review where all eyes are required to be on the documents or where a coverage review is required, ie a review The decision as to what model to rely on is determined on what is required of the document set, whether all eyes are required on all documents before they go out the door and to whom the documents are being produced (think litigation versus regulator). Whether the determining factor as to how you choose to use AI Review is dependent on resourcing on case requirements, it is a well known and defensible approach to review.
Gen AI
Gen AI will be a feature of review going forwards. The models are just being used to help save time and Law In Order are able to discuss how it can be used in your matters. The major feature of Gen AI will in how to craft prompts and how to validate results.
Conclusion:
Technology can be used to speed review and assist with consistency. It starts with planning, careful collection and knowledge of the appropriate technology to apply. To know more about how eDiscovery technology can benefit your legal team, talk to one of our eDiscovery experts today! @Contact
For further information, please contact:
Murali Baddula, Chief Digital Officer, Law In Order
sydney@lawinorder.com
