Gilbert + Tobin and the Push to Rework Legal Work With AI

Gilbert + Tobin, one of Australia's well-known commercial law firms, has become the subject of an OpenAI customer story about changing how a modern legal practice is run. The piece, centered on the firm's Sam Nickless, frames AI less as a gadget for individual lawyers and more as a lever on the business model of law itself.

Why law keeps showing up in AI case studies

Legal work is an obvious target for generative AI. It is built on text: contracts, disclosure documents, correspondence, case summaries and regulatory filings. Much of it is repetitive, time-pressured and expensive, and a large share of the effort goes into reading, comparing and drafting rather than into the judgment that clients actually pay a premium for.

That makes the sector a natural showcase for vendors, and OpenAI is far from alone in courting it. Harvey, which has raised substantial venture funding and counts several large firms as customers, built an entire company around legal AI. Thomson Reuters and LexisNexis have layered assistants into their research products, and a crowd of smaller startups targets niches such as contract review and due diligence. Large firms have been running pilots, forming internal innovation teams and negotiating enterprise agreements with model providers for the past couple of years.

The business-model question

The more interesting thread in a story like this one is not the tooling but the economics. Law firms have traditionally priced work by the hour. If AI compresses a task that took a junior associate a day into something that takes an hour, the billable-hours model starts to look awkward. Firms face a choice: absorb the efficiency, pass the savings to clients, or move toward fixed fees and value-based pricing that reward outcomes rather than time spent.

There is also a structural effect on talent. Routine document review has long served as a training ground for junior lawyers. If that work is increasingly automated, firms need new ways to build skills and judgment in early-career staff. Framing AI adoption as a rethink of the business, rather than a tool rollout, suggests firms like Gilbert + Tobin are at least thinking about these second-order effects.

What it means for developers

For engineers, the legal market is a demanding but lucrative proving ground. The requirements are stricter than in most enterprise settings:

• Accuracy and traceability. Lawyers need to verify every claim, so systems that cite sources and expose their reasoning have an edge over those that merely produce fluent text.
• Confidentiality. Client privilege and data-residency rules shape architecture, from where models run to how documents are retained.
• Integration. Value comes from fitting into document management systems, matter workflows and review processes, not from a standalone chat box.
• Evaluation. Firms want measurable benchmarks on their own work product, which pushes vendors toward serious testing rather than demos.

The open-source angle matters here too. Some firms and legal-tech teams prefer open-weight models they can host themselves for confidentiality reasons, and a growing set of retrieval, evaluation and document-parsing projects supports that approach. Proprietary vendors compete on convenience and capability; self-hosted stacks compete on control.

What to watch next

The signals worth tracking are practical ones. Do firms move away from hourly billing in meaningful numbers? Do clients start demanding AI-driven efficiency as a condition of engagement? How do regulators and professional bodies treat AI-assisted work, including questions of supervision and liability when a model gets something wrong? And do the quieter infrastructure choices, such as hosted versus self-managed models, start to differentiate firms from one another?

Customer stories are marketing by nature, and they rarely show the friction of real deployments. Still, a prominent firm publicly describing AI as a reason to rethink its operating model is a useful marker of where the profession is heading. For developers, the message is that the opportunity in law lies less in clever prompts than in dependable, auditable systems that fit how legal work is actually done.