Gilbert + Tobin rethinks the business of law with AI
Nov 11, 2026, 4:00 PM · OpenAI

An Australian law firm's chief executive says AI's payoff so far is time, not headcount, and the way he got there is a better template than most vendor case studies.
Why it matters
Law is one of the professions where AI anxiety runs highest: confidential client data, billable hours, and junior work that looks easy to automate. Gilbert + Tobin, one of Australia's leading corporate firms, is now on the record moving from individual chatbot use to multistep agentic workflows on ChatGPT Enterprise, ChatGPT Work and Codex, with people still responsible for review and approval.
The number that stands out is adoption: the firm says 87% of enabled users are active, more than twice what it typically sees for other tools, and it got there without usage mandates.
From the desk
We'll flag the obvious first. This is an OpenAI-published interview with a customer, so it's a showcase, not an audit. There are no figures here on error rates, client outcomes or billing changes. Read it as one leader's account of how he rolled the tools out.
That said, the rollout is the useful part, and we think it's sound. The firm started with open experimentation and a hard rule against putting client or confidential data in. It then moved to an enterprise deployment with Australian data residency, role-based access and admin controls before scaling. And the cultural message was blunt: AI is not cheating, but the person still defines the task, checks the output and owns the result. That's the right order of operations for any regulated profession.
The detail I keep coming back to is the CEO's digital twin, a custom GPT trained on his writing, feedback he's received and how he works, which his direct reports now use to pressure-test ideas before they bring them to him. He's careful to say it doesn't make decisions or speak for him. Useful, yes. But it's also a preview of a workplace where people rehearse for the boss with a model of the boss. If that becomes normal, organizations risk sanding off disagreement before it ever reaches the room.
Then there's the annual review example: confidential material for 120 partners, assembled in minutes, at a usage cost he put at about $62, for work he says would otherwise have taken several people multiple days. He frames AI's value as freeing staff for the more human side of the job rather than cutting people. We'd like to believe that, and it may hold at a firm that's growing. The harder question is what happens to the junior work that used to train young lawyers when tasks like this vanish into a dashboard.
Context
Gilbert + Tobin works across capital markets, M&A, disputes and technology. The firm describes its path in three phases: experimentation, personal productivity and organizational productivity. Leaders used the tools themselves, and internal showcases, challenges and an early AI bounty brought ideas in from across the firm.
Who feels it
- Law firms
- A practical rollout pattern: constrain data early, secure the enterprise channel, then push leadership use over mandates.
- Junior professionals
- Admin and assembly work that once built skills is collapsing into minutes; firms will need deliberate ways to train judgment.
- Clients
- Faster turnaround is plausible, but clients should ask how review and accountability work when agents draft the material.
What to watch
- Whether the firm or its peers publish measurable quality or error data, not just adoption rates
- How firms handle pricing and billable-hour pressure as agentic workflows spread
- Whether digital-twin tools for executives spread, and what governance rules they come with