Clayton Utz

Legal research in large law firms is an inherently labour-intensive process. Attorneys must manually search across vast repositories of legislation, case law, statutes, and secondary commentary — work that is both time-consuming and cognitively demanding. At enterprise scale, this inefficiency has a compounding cost: billable hours consumed by rote research are hours unavailable for higher-value client advisory work. A Stanford University study found that generative AI legal research tools produce false or misleading information between 17% and 33% of the time, meaning firms could not simply adopt generic AI — they needed purpose-built tooling with robust safeguards before deployment was viable.

Clayton Utz adopted Lexis+AI, a purpose-built legal research assistant from LexisNexis, built on Anthropic's Claude 2 and OpenAI's GPT-4 and trained on LexisNexis' repository of 1.23 million court opinions, statutes, filings, and secondary materials. The platform provides conversational search, case summarisation, citation validation, and intelligent document drafting. Critically, it addresses the two primary risks of generative AI in legal practice: client confidentiality (prompts are encrypted, not used for model training, and deleted after use) and hallucinated citations (every generated document has its citations validated against the full LexisNexis primary content database before delivery). Clayton Utz participated in co-development and testing prior to full production adoption, making it the first Australian law firm to deploy the solution for day-to-day legal work.

The productivity gains were concrete and immediate. One documented example from co-adopter Holding Redlich: a research task that took a lawyer 4.5 hours using traditional methods was completed in 30 minutes with Lexis+AI — a substantial reduction in research time. Qualitative outcomes included:

• Lawyers able to quickly identify relevant cases and discard irrelevant ones via the case summary feature
• Faster document drafting with built-in citation validation reducing manual verification burden
• Clayton Utz positioned as the first Australian firm to deploy the technology at scale, signalling a firm-wide commitment to AI-enabled practice

Clayton Utz Chief Executive Partner Emma Covacevich noted the deployment unlocked "efficiencies that haven't previously been possible."

• Citation validation is non-negotiable: Legal AI tools must verify generated references against authoritative databases — hallucination rates of 17–33% in generic models make this a hard requirement, not a nice-to-have.
• Confidentiality architecture matters: Ensure prompts are encrypted, not used for model training, and deleted post-session before deploying any AI tool in a client-facing legal context.
• Pilot participation accelerates adoption: Clayton Utz co-developed the product before rollout, which reduced deployment friction and built internal confidence in the tooling.
• AI augments, not replaces, legal judgment: Lawyers must still read cases, verify citations, and apply professional due diligence — the tool compresses research time, not professional responsibility.