Holding Redlich will trial drafting legal documents and emails to clients on ‘Lexis+ AI’ before the generative AI platform’s commercial release in the Asia Pacific.
LexisNexis tapped early adopters of its previous solution ‘Lexis+,’ which automates lower-risk tasks, to provide feedback on the upgrade’s performance generating court filings, advice to clients and other documents within the Australian legal sector, to also sign on to trial AI.
The law firm’s chief knowledge officer Keren Smith told iTnews that after a three-month trial using Lexis+ to automate research tasks, the firm was confident about deploying Lexis+ AI — currently released only in the US — to expand into new territory.
“We want our staff to have the best resources and tools available, not only to provide our clients with timely and excellent service but also to attract and retain the best staff,” she said.
In October, Holding Redlich hosted a Turing test that confirmed Lexis+’s ‘argument analyser’ feature could outperform a human at strengthening a lawyer’s submission by suggesting relevant case law to reference in it.
The law firm has a longstanding partnership with the LexisNexis, best known for its court documents subscription services.
Smith said she was excited for the company’s 70 partners and roughly 500 staff to pilot Lexis+ AI’s extended list of capabilities.
Built on two LLMs — anthropic’s Claude 2 and OpenAI’s GPT-4 — and trained on LexisNexis’ 1.23 million court opinions, statutes, filings, and secondary materials, Lexis+ AI can automatically generate documents.
Lawyers enter prompts and materials that help contextualise what they need Lexis+AI to write and it returns first drafts of documents like advice to clients, internal emails or court filings, such as statements of claim.
Keeping a human in the loop, the user is then advised to review the output and use prompts to tweak its language to the intended tone.
LexisNexis Asia Pacific managing director Greg Dickason told iTnews that the platform also has safeguards against the two greatest risks of applying generative AI to the legal sector: infringing client confidentiality and producing legal documents that reference inexistent statutes or case law.
“Prompts are encrypted and the model does not learn from them. They are also deleted so IP is protected,” he said.
Lexis+ AI also mitigates against the risk of generating court fillings containing hallucinated precedents.
After generating a court document, Lexis+ AI validates its citations against Shepard’s and LexisNexis’ other proprietary citation indexes.
In addition to its document generation capabilities, Lexis+ AI has a ‘document upload’ and ‘summarisation’ feature, which will enable Holding Redlich to add to the list of research tasks it has tested automating since deploying Lexis+.
Holding Redlich’s lawyers, who currently review large documents manually, will trial uploading them to Lexis+ AI, which can extract their key points and translate them to brief descriptions.
“Holding Redlich’s team has completed a thorough trial of Lexis+ and are …anticipating the significant legal research efficiencies that Lexis+AI will bring,” Smith said.
Lexis+’s research assistance capabilities include automated comparisons of current and previous legislation to highlight changes over time and a feature that responds to users’ natural language search queries about case law to find answers more efficiently.
LexisNexis also tapped Clayton Utz to trial Lexis+ AI; other large law firms investing in generative AI include MinterEllison and Lander & Rogers, which have deployed Microsoft Copilot.