Tuesday, 18 June 2024

Legal discovery is primed for AI disruption.

Incremental technical innovations have driven continuous improvement in litigation discovery for some time, but the exponential growth in data volumes and complexity, along with rapid advances in generative AI, have created new urgency for the legal industry to consider AI-powered document review options.   

Discovery requires efficiency, transparency, and fairness without making access to justice prohibitively expensive. This can be an overwhelming challenge when faced with wide-ranging and complex discovery categories and huge quantities of documentary evidence – especially when the cost of manual review continues to increase, and existing ‘advanced’ technology solutions can evaluate documents for only a single issue at a time. These factors mean AI technology will soon be an essential part of almost all discovery procedures.  

KordaMentha has been given advanced access to Relativity aiR for Review, a powerful generative AI tool set designed specifically for use in discovery procedures. Properly deployed, this new technology allows a smaller team of subject matter experts to easily outperform larger review teams using traditional methods in efficiency, cost-effectiveness and coding accuracy. As the only RelativityOne Certified Partner and consulting firm in Australia with access to aiR for Review through its limited availability program, KordaMentha is a pioneer in the adoption of this powerful new technology, which expedites the often consuming and resource-intensive discovery process.   

What is aiR for Review and what value does it provide?   

Conventional review methods mean reviewers must sift through huge volumes of often irrelevant data in their search for pertinent documents. Moreover, it is challenging to gauge the probable number of relevant documents within a particular set until the review process is well underway, complicating estimations of resource allocation and timing. aiR for Review alleviates these challenges by isolating relevant material so significantly fewer documents require manual inspection.   
aiR for Review analyses large volumes of documents with unparalleled speed and accuracy. Backed by Relativity’s and Microsoft’s leading security postures, aiR for Review is deployed in a secure and restricted environment, maintaining confidentiality of client data while allowing the quick and cost-effective identification of relevant documents.   
Instead of relying on a large and expensive document review team with varied availability, experience, and interpretations, aiR for Review empowers trusted subject matter experts to guide the AI’s analysis with carefully crafted prompt criteria of up to 10,000 characters. This set of instructions, which details the ‘who’, ‘what’, ‘when’ and ‘why’ at the core of the case, is relied upon by aiR for Review to categorise documents as ‘relevant’ or otherwise. A smaller team of senior reviewers can then adopt or overturn this proposed categorisation, as with any standard second-line review, to ensure human stakeholders retain control of the review process.   
Unlike existing solutions or manual review workflows, aiR for Review leverages generative AI to provide a written rationale explaining the conclusion it has reached for each document, meaning your experts can refine and iterate on their prompt criteria until they and the AI are on the same page. In addition, aiR for Review provides counter-considerations, and cites supporting passages from within the document to ensure maximum transparency of its robust reasoning.    

The result? Rather than ineffective keyword searches or large knowledge-diluted review teams, a smaller group of trusted advisors can partner with AI to directly influence the coding of every document, improving not only efficiency, but also coding accuracy.    


Figure 1 – Relevance review of 20,000 documents against two distinct issues, where 2,000 documents are relevant   

 aiR for Review can also be paired with other functionality more familiar to Australian lawyers, such as TAR and email threading, with bespoke workflows tailored by eDiscovery experts amplifying its capabilities even further. Whether a legal team is preparing its own discovery, responding to a regulatory notice, running an investigation, interrogating discovery provided by another party, or undertaking any of the other established use cases for the application, AI presents unprecedented opportunities and potential.   

Why AI-powered discovery must be embraced   

No area of litigation is as primed for AI disruption as discovery.    
Discovery is a proportionate exercise that needs to facilitate efficiency, transparency, and fairness without making access to justice prohibitively expensive. Whilst innovation has seen approaches to discovery mature continuously, the exponential growth in data volumes and data complexity makes AI-powered document review options, such as aiR for Review, essential for maintaining this crucial balance.    

The Victorian Supreme Court, a long-standing advocate for technology as the first in Australia to order the use of Technology Assisted Review (TAR), seems to agree. In its ‘Guidelines for the Responsible use of Artificial Intelligence in Litigation’ released in May 2024, the first application case proposed specifically deals with the document review and discovery processes:   

“AI that can search and identify relevant matters in a closed category of information is presently used with good effect in the Court. An illustration is the use of Technology Assisted Review (TAR) which employs machine learning for large scale document review. Practitioners should consider the use of such options to improve productivity and efficiency consistent with the expectation that use of common technologies is a core skill for lawyers. Use of such technology with co-operation of the parties and appropriate management by the Court is to be encouraged.”   

Onus placed on lawyers   

While not proscriptive on specific technology, the principles set out by the Victorian Supreme Court are nonetheless clear in terms of what practitioners should expect. Being that:   

  • Courts will look to be proactive – AI will be pervasive in our legal system, whether Courts embrace it or not. Australia and the UK were four years behind the US when embracing TAR, but this time will look to move more quickly. The technical literacy and documented benefits of TAR will facilitate this transition and will ensure parties are able to deploy cutting-edge solutions if they can demonstrate they have been used appropriately.    
  • AI use must be responsible and defensible – since it was endorsed in 2016, Australian courts, regulators, and parties in dispute have developed an appreciation for the ability to validate the completeness of a TAR process statistically and objectively. The Supreme Court’s recent guidance suggests this should not change just because AI is involved. Parties will still need to demonstrate their discovery process, AI-driven or otherwise, is defensible and enables practitioners to meet their various obligations. Recent decisions, including one in which KordaMentha provided evidence to establish another party had not discharged their discovery obligations, point to the Court’s willingness to interrogate this issue in detail.   
  • Lawyers should understand AI applications they rely on – it is incumbent on lawyers to leverage technologies that are beneficial to their clients and the court. Especially as corporations and firms begin to deploy AI in closed proprietary applications, it can be difficult for other participants to a proceeding to interrogate their output. This renders more broadly accessible solutions known to the court and to the respective parties as preferable in fostering this understanding. Even so, whilst applications like aiR for Review present incredible opportunities, leveraging them without appropriate workflows and expert guidance can lead to problems.    
  • Current expectations and limitations will be challenged – TAR enabled parties to consider millions of documents with groundbreaking efficiency. aiR for Review goes a step further, not least because it is not binary like TAR and can classify documents against multiple categories, saving exponential amounts of time on multi-issue disputes. For law firms without dedicated document-review teams, this feature will enable them to work with clients and on engagements previously considered unmanageable. It will also provide clients with greater choice when seeking representation, allowing them to retain legal expertise without being limited to only those practitioners supported by extensive review resources.    

The risks presented by AI are most likely to be realised when it is used incorrectly, inappropriately, or without consideration. As such, AI should not be considered a replacement for a lawyer’s due care and skill. Instead, aiR for Review can amplify a lawyer’s impact on a case, doing the ‘heavy lifting’ to increase the efficiency and effectiveness of document review processes in an area of litigation that can only benefit from continued, responsible innovation.   

When combined with aiR for Review and KordaMentha’s expertise, practitioners can be assured that:    

  • algorithms relied upon have been developed responsibly and ethically    
  • client data remains secure and confidential   
  • the underlying technology is understood, reliable, and deployed in an informed manner; and   
  • outputs can be validated to an appropriate and defensible standard   

This being the case, the case for AI in discovery is not ‘if’ or even ‘when’, but ‘now’.