Gilbert + Tobin expands legal AI use with ChatGPT and Codex

Australian corporate law firm Gilbert + Tobin has expanded its use of OpenAI products from early experimentation to structured organisational workflows. The firm uses ChatGPT Enterprise, ChatGPT Work and Codex, while keeping people responsible for reviewing and approving work. It reports that 87% of enabled ChatGPT users are active, a rate it says is more than twice that typically seen for other tools.
Sam Nickless, Partner and Chief Executive Officer at Gilbert + Tobin, said the firm is pursuing three objectives: better client services, a stronger workplace for employees and leadership in its market. He said the value identified so far is not primarily lower cost or fewer staff, but giving people more time for client meetings, coaching and mentoring.
From open experimentation to controlled deployment
When ChatGPT first became available, Gilbert + Tobin encouraged staff to test possible uses in a constrained setting. The firm made clear that client data and confidential information could not be entered. That approach was designed to let employees explore where the technology could be useful while respecting the demands of legal practice.
The next phase introduced ChatGPT Enterprise in a controlled environment. Nickless said Australian data residency, contractual protections, role-based access and administrative controls gave the firm confidence to broaden use while meeting its own requirements and those of clients. The firm frames governance as an enabler of responsible use rather than a barrier to adoption.
Its instruction to staff is that AI is another tool for applying professional judgement, not a substitute for it. Users must define the task, check the output and accept responsibility for the final result. That distinction is particularly relevant in a firm handling work across capital markets, mergers and acquisitions, disputes and technology.
Leadership and practical workflows
Nickless said senior leaders need to use the technology directly, experiment with its limits and show employees that learning matters. Gilbert + Tobin did not set mandatory usage targets. Instead, its business transformation team demonstrated workflows suited to different functions, supported by videos, showcases, challenges and an early AI bounty intended to surface ideas from throughout the firm.
For his own work, Nickless built a custom GPT described as a “digital twin”, using examples of his writing, received feedback, firm context and his preferred working style. The tool is now moving to Plugins. His direct reports can use it to test their thinking against that context before meeting with him; it does not make decisions or speak on his behalf.
He also used ChatGPT Work during a confidential annual review process for the firm’s 120 partners. The product updated files with spreadsheet data, created PDFs and bundled completed materials in minutes. Nickless said the usage dashboard put the cost of that work at about $62, compared with an effort that otherwise would have involved several people over multiple days.
What businesses can take from the case
Gilbert + Tobin’s model combines executive participation, clear restrictions on sensitive information, administrative controls and practical examples for specific roles. For businesses considering wider AI deployment, the implication is to begin with real work, establish the conditions for secure testing and retain accountable human review as workflows become more complex.

