Everything About AI
NextFin News - At Gilbert + Tobin, one of Australia's leading corporate law firms, a conflict, know-your-customer and anti-money-laundering check that once consumed three to eight hours of professional time now finishes in about five minutes. A recruitment research workflow that took roughly four hours is done in 20 minutes. Preparing audit reports across 300 entities no longer takes a day of manual labour. These are measured results from a firm that has moved artificial intelligence from experimentation into the core of how it operates: as of June 2026, 87% of enabled ChatGPT seats were active, more than twice the adoption rate the firm typically sees for other tools. The question the case raises for the wider professional-services industry is not whether AI can save time, but whether a profession built on the billable hour can survive the time it saves.
The Situation: AI Enters the Operating Model, Not the Advice
Gilbert + Tobin, a firm spanning capital markets, mergers and acquisitions, disputes and technology, has taken an unusual path into AI. Rather than promising to automate legal advice, the firm has trained its deployment on the operational work that surrounds it: recruitment, finance, marketing, business development, conflict checks, KYC, AML screening and audit reporting. The distinction is deliberate. The firm's public position is that AI should not change how it delivers legal advice, but should raise the standard of the work that supports that advice.
The rollout followed three phases. First, employees were encouraged to experiment with AI under a strict rule: no client data or confidential information was to be entered. That created a safe, constrained environment to learn what the technology could do. Second, the firm introduced ChatGPT Enterprise in a controlled environment and expanded from there, with Australian data residency, contractual protections, role-based access and administrative controls giving leadership the confidence to scale. Third came the shift toward organisational productivity, using ChatGPT Work and Codex to build defined, multistep agentic workflows in which people remain responsible for review and approval.
The adoption numbers are the headline. As of June 2026, 87% of enabled ChatGPT seats were active. The firm did not mandate usage targets; instead it invested in role-specific enablement, with the business transformation team joining individual team meetings to demonstrate workflows tailored to each function. CEO Sam Nickless introduced the technology through examples from his own work and a deliberately simple message: "AI is not cheating." The phrase reframed the tool from an inappropriate shortcut into another way of applying professional judgment.
The quantified results, published in OpenAI's customer case study, are:
- Recruitment research and data extraction: approximately four hours down to about 20 minutes.
- Selected conflict, KYC, AML and politically exposed person checks: three to eight hours down to five minutes.
- Audit report preparation across 300 entities: one day of manual work avoided.
- A separate reference-processing workflow saving an estimated 25 minutes per candidate.
The firm has also applied Codex to a previously laborious component of appraisals for its 112 partners, and in September 2026 announced it had been named a finalist in the AI Innovator - Legal Practice category at the Australian AI Awards 2026.
The Mechanism: Automating the Substrate, Retaining the Judgment
The first-order reading of these numbers is simple productivity. The deeper mechanism is that AI is being applied to the substrate of professional work - the research, extraction, cross-checking and assembly tasks that consume the bulk of associate and operations time - while the firm deliberately retains the judgment layer. This is a separation of the commodity from the craft.
Aviva Leitch, the firm's Head of Business Transformation, put the distinction plainly:
"Codex can take AI from being a helper to being a doer. It carries out the steps in an operational workflow, while our people remain responsible for reviewing and approving the result."
- Aviva Leitch, Head of Business Transformation, Gilbert + Tobin
That sentence captures the operating model: the workflow is automated end to end, but accountability stays human.
The transmission channel is the billable hour. Law firms sell time; their revenue model prices specialist attention by the six-minute unit. When a four-hour task becomes a 20-minute task, the firm faces a choice: bill fewer hours for the same output, or redeploy the freed time into higher-value work. Gilbert + Tobin has publicly chosen the second path. Nickless has said the value seen so far has not been about reducing cost or needing fewer people, but about freeing people to focus on the human side of the work - meeting clients, coaching teams, mentoring future leaders. As he put it in an interview, the goal has been "freeing them up to then do the more human side of their work."
That is a coherent position for a firm that sells judgment. It is also a position under pressure, because the time saved does not disappear from the market; it becomes available to competitors.
Cyclical Wave or Structural Shift: Why This Time Is Different
This is the decisive question. A cyclical productivity wave - the kind law firms have absorbed before through document management systems, legal research databases and outsourced discovery - eventually reverts or gets absorbed into the existing pricing model. A structural shift changes the rules permanently.
Three pieces of evidence point to structural rather than cyclical.
First, the capability is not a marginal improvement in speed; it is a change in what the tool can do. The move from ChatGPT as an assistant to Codex as an agent that executes defined workflow steps means the technology now performs multi-step work, not just drafts text. An agent that researches, extracts, cross-references and assembles a report is replacing a process, not a keystroke. Processes do not revert.
Second, the economics of the profession are being inverted. The traditional law-firm pyramid depends on large cohorts of junior lawyers doing precisely the kind of research and review work that AI now performs fastest and cheapest. When the bottom of the pyramid is automated, the training model, the leverage model and the path to partnership all change. That is not a cycle; it is a redesign of the firm's internal labour market.
Third, the trust infrastructure required for adoption has finally arrived. Gilbert + Tobin's rollout was gated on Australian data residency, contractual protections, role-based access and administrative controls. Earlier generations of AI could not have been deployed in a regulated professional-services environment with this level of confidence. The gating factor was not demand - it was governance. That gate has now opened.
The counter-argument has real force: this is an efficiency wave that will be priced back into the billable hour. Clients have long resisted moving from time-based to value-based pricing, and law firms have proven remarkably resilient in preserving their revenue models through previous technology shocks. On this view, the time saved becomes slack or gets reabsorbed into more thorough work rather than lower fees.
That argument holds in the short term. But it underestimates the second-order effect: competition. Once one firm can deliver the same assurance work in five minutes instead of five hours, a rival bidding for the same mandate faces a choice between matching the efficiency or absorbing the cost disadvantage. Efficiency advantages in professional services do not stay private for long; they become the market's new cost base. The firm that refuses to adopt does not preserve its margins - it loses the work.
The falsifying signal for the structural thesis is specific and observable. If, within two years, major Australian corporate firms report stable or rising revenue per lawyer alongside flat utilisation of AI tools in operational workflows, and if client fee arrangements remain predominantly time-based with no shift toward fixed-fee or outcome-based pricing for the automated work categories, then the technology has been absorbed into the old model rather than replacing it. Conversely, if fixed-fee pricing becomes standard for conflict, KYC and compliance work and leverage ratios fall, the structural shift is confirmed.
The Second-Order Question: Who Owns the Margin?
The conventional read is that AI adoption benefits the firm that adopts it. The second-order question is whether the benefit accrues to the firm, the client, or the technology vendor.
In the short run, the adopting firm captures the margin: it can underbid competitors, improve turnaround, or redeploy staff. In the medium run, competitive pressure passes the savings to clients through lower fees or fixed pricing. In the long run, the persistent beneficiary may be the platform provider. If the operational substrate of legal work runs on a small number of AI platforms, the economics of the profession begin to resemble a tollbooth model: the value created by automation is partially captured as subscription, usage and token costs paid to the vendor.
This is not hypothetical for Gilbert + Tobin. The firm's confidence to expand access rested explicitly on moving to an OpenAI environment with Australian data residency. Chief Information Officer Mitch Owens said:
"OpenAI approached the platform from an enterprise perspective and built the controls that organisations like ours need to operate with confidence."
- Mitch Owens, Chief Information Officer, Gilbert + Tobin
That statement is both a governance endorsement and a dependency. The firm has tied its operating model to a vendor's roadmap, pricing and data-residency commitments.
The Counter-Thesis, Taken Seriously
The strongest argument against the transformation narrative comes from the profession's own incentives. Law firms are partnerships; their economics reward realisation and leverage. AI that makes junior work faster threatens the apprenticeship model that produces partners. It also creates liability risk: a hallucinated citation or a missed conflict in an automated workflow can produce malpractice exposure that no efficiency gain offsets.
Nickless is aware of this. He has acknowledged the huge risks involved if the firm's lawyers rely on hallucinating artificial intelligence models to help build their cases, while remaining committed to using the tools to run the practice. The firm's answer is governance plus human review: approved-task lists, guidance on what may be entered, output review requirements, and the principle that people remain accountable for the result.
This is a defensible position, but it contains a tension. Human review is only as good as the reviewer's ability to detect errors. If AI does most of the work and a human reviews the output, the reviewer faces the same automation bias that affects pilots, radiologists and auditors: the tendency to trust the machine's output. The firm's governance model assumes review is meaningful; whether it remains meaningful at high automation levels is an open empirical question, not an assumption.
What Comes Next
The Gilbert + Tobin case is a template for how a knowledge-work business can adopt AI without promising to replace its core professional judgment. The approach - start with experimentation under constraints, build trust through governance and data residency, demonstrate role-specific value, let adoption follow rather than mandating it - is replicable across professional services: accounting, consulting, insurance, and the compliance functions of large corporations.
The short-term impact is operational: faster turnaround, lower internal cost per matter, and freed capacity. The medium-term impact is competitive: firms that do not build comparable workflows will find themselves bidding against firms that can perform the same assurance work at a fraction of the time cost. The long-term impact is structural: if the substrate of professional work is automated, the billable hour cannot survive unchanged, the leverage pyramid flattens, and pricing shifts toward value and outcomes.
Three signals are worth watching. First, pricing: whether conflict, KYC and compliance work moves to fixed fees. Second, leverage: whether the ratio of junior to senior lawyers falls as entry-level work is automated. Third, vendor concentration: whether a small number of AI platforms become the operating system for professional services, capturing a growing share of the value created.
The base case is that adoption spreads across the top tier of Australian and global firms over the next two to three years, with efficiency gains passed partly to clients and partly retained as margin during the transition. The upside case for the profession is that freed capacity expands the market - cheaper compliance and transaction work brings in clients who previously priced themselves out. The downside case is a painful compression of the traditional partnership model, with fewer junior roles and a prolonged debate over how to train lawyers when the apprenticeship work has been automated.
Gilbert + Tobin's real innovation is not the technology. It is the decision to apply AI to the business of law rather than to the practice of law - to automate the substrate while keeping judgment human. That distinction is what makes the adoption stick, and what makes the threat to the old model structural rather than cyclical. The firms that survive will be the ones that figure out how to sell judgment when the hours behind it have disappeared.
Explore more exclusive insights at nextfin.ai.
