Gilbert + Tobin says 87% of enabled staff use ChatGPT, without usage targets

Australian law firm Gilbert + Tobin says 87% of its enabled ChatGPT users are active—more than twice its usual adoption rate for other tools. In an OpenAI customer interview, CEO Sam Nickless describes an approach built around leaders using AI themselves, examples tailored to employees’ jobs and human responsibility for the finished work.

The figure measures activity among people given access, not the share of the entire firm using ChatGPT. It is a customer-reported result published by the supplier, rather than an independently verified finding. Nickless says the firm did not mandate usage targets; it worked to make the technology relevant.

From safe experiments to shared workflows

Nickless describes three stages: experimentation, personal productivity and organizational productivity. Early on, employees could explore AI, but could not enter client data or confidential information. The firm then introduced ChatGPT Enterprise in a controlled environment, before broadening its use of OpenAI tools.

He credits Australian data residency, contractual protections, access tied to employees’ roles and administrative controls with giving the firm confidence to scale. OpenAI says the deployment now includes ChatGPT Enterprise, ChatGPT Work and Codex, moving beyond individual assistance into defined workflows with multiple steps and human review and approval.

The adoption effort gave employees several routes into the technology, rather than relying on an instruction to use it:

• The business transformation team demonstrated workflows tailored to different functions.
• Videos, showcases and challenges helped surface ideas across the firm.
• An early AI bounty added another way to draw out employees’ suggestions.

Nickless says the value so far has not been reducing costs or needing fewer people. Instead, he points to freeing employees for client meetings, coaching and mentoring. His message to staff is that using AI is not cheating, but neither does it transfer responsibility away from the person using it.

They still need to define the task, check the output and take responsibility for the final work.

Sam Nickless, partner and CEO, Gilbert + Tobin, in OpenAI’s interview

Rehearsing the CEO’s feedback

One of Nickless’s own experiments is a “digital twin”: a custom GPT supplied with examples of his writing, feedback he had received, information about the firm and his working preferences. He says it began as a personal tool and became more useful when he shared it with his direct reports.

Team members can use the tool to test their thinking against his context and style before approaching him. Nickless draws a boundary around that simulated feedback: the system does not make decisions or speak for him. Its purpose is to give the eventual conversation a stronger starting point, not replace the conversation.

A partner-review task with a visible bill

Nickless also describes using ChatGPT Work for a confidential annual review process involving the firm’s 120 partners. He asked it to update files from spreadsheet data, generate PDFs and bundle the finished material. He says the work finished in minutes; checking the usage dashboard afterward showed a cost of about $62.

He judged that a good trade for a task he says would otherwise have involved several other people over multiple days. Alongside that example, he stresses the need for administrative controls and detailed usage information. His advice to other CEOs combines firsthand experimentation with clear rules, practical examples and people remaining accountable for results.