Australian law firm Gilbert + Tobin adopts ChatGPT Enterprise to build agentic workflows
Article Summary
Australian law firm Gilbert + Tobin has integrated ChatGPT Enterprise and Codex to transition from individual AI assistance to defined, multistep agentic workflows overseen by human staff. The firm reports an 87% active usage rate among enabled employees, which the company states is more than double the typical adoption rate for its other internal tools. Under this initiative, personnel retain responsibility for defining tasks and verifying all AI-generated outputs while utilizing the platform within a governed, secure environment that mandates data residency and administrative controls.
AI Analysis
## Core Assessment
Gilbert + Tobin's rollout of ChatGPT Enterprise demonstrates a realistic path for deploying large language models in professional services: frame the system as an augmentative tool rather than an automated replacement, enforce strict perimeter controls, and hold human staff legally and operationally accountable for final output. The reported 87% active usage rate among enabled employees indicates strong initial engagement, driven largely by executive modeling and role-tailored workflow demonstrations rather than top-down mandates. By insisting on Australian data residency, role-based permissions, and explicit verification steps, the firm addresses the primary enterprise barrier to generative AI—confidentiality risk—while moving beyond basic text generation into structured, multistep task execution.
However, moving from ad-hoc productivity gains to automated multistep pipelines introduces subtle operational bottlenecks that marketing case studies rarely highlight. Even with administrative controls and administrative cost tracking—such as processing partner review documents for $62—the economic and operational bottleneck shifts entirely to the verification phase. When AI systems generate dense work products in seconds, human reviewers face cognitive fatigue checking citations, spreadsheet joins, and narrative nuances. Unless organizations measure the actual time and liability cost of human auditing, high internal adoption metrics can obscure whether overall operational risk and review overhead have actually decreased.
## Impact and Advice
**Legal Operations Directors:**
Establish verifiable review pipelines before scaling multistep workflows. Enforce explicit audit logs indicating who checked each generated document section, ensuring staff understand that productivity gains disappear if unverified hallucinations slip into client-facing deliverables or confidential partner workflows.
**Enterprise IT and Security Architects:**
Prioritize data tenancy and role-based access governance over broad feature rollouts. Audit tenant administrative logs regularly to verify that data residency boundaries hold firm and that custom internal bots or executive prototypes do not expose restricted management context across different organizational tiers.
## Bottom Line
Getting lawyers to log into an AI portal is easy; proving that humans are diligently auditing those high-speed outputs before they trigger professional liability is the real engineering challenge.
Gilbert + Tobin's rollout of ChatGPT Enterprise demonstrates a realistic path for deploying large language models in professional services: frame the system as an augmentative tool rather than an automated replacement, enforce strict perimeter controls, and hold human staff legally and operationally accountable for final output. The reported 87% active usage rate among enabled employees indicates strong initial engagement, driven largely by executive modeling and role-tailored workflow demonstrations rather than top-down mandates. By insisting on Australian data residency, role-based permissions, and explicit verification steps, the firm addresses the primary enterprise barrier to generative AI—confidentiality risk—while moving beyond basic text generation into structured, multistep task execution.
However, moving from ad-hoc productivity gains to automated multistep pipelines introduces subtle operational bottlenecks that marketing case studies rarely highlight. Even with administrative controls and administrative cost tracking—such as processing partner review documents for $62—the economic and operational bottleneck shifts entirely to the verification phase. When AI systems generate dense work products in seconds, human reviewers face cognitive fatigue checking citations, spreadsheet joins, and narrative nuances. Unless organizations measure the actual time and liability cost of human auditing, high internal adoption metrics can obscure whether overall operational risk and review overhead have actually decreased.
## Impact and Advice
**Legal Operations Directors:**
Establish verifiable review pipelines before scaling multistep workflows. Enforce explicit audit logs indicating who checked each generated document section, ensuring staff understand that productivity gains disappear if unverified hallucinations slip into client-facing deliverables or confidential partner workflows.
**Enterprise IT and Security Architects:**
Prioritize data tenancy and role-based access governance over broad feature rollouts. Audit tenant administrative logs regularly to verify that data residency boundaries hold firm and that custom internal bots or executive prototypes do not expose restricted management context across different organizational tiers.
## Bottom Line
Getting lawyers to log into an AI portal is easy; proving that humans are diligently auditing those high-speed outputs before they trigger professional liability is the real engineering challenge.