Eight and a half minutes

That's how long it took Claude, the AI model my team and many others now use every day, to complete a University of Melbourne defamation law exam that human students had three days to write.

It scored 70.25%. The student average was 74%. Both sit comfortably in the same honours band.

I spent yesterday at the Law Institute of Victoria's AI Forum at Federation Square. The exam result was shared by Dr Matt Collins AM KC, who teaches the subject and ran the experiment himself, as a lecturer at the University of Melbourne. He created a fake student number, submitted Claude's paper alongside the real ones, and had both marked blind by two independent assessors. He's a sitting silk testing his own field. The result wasn't a parlour trick from a demo or a piece of generated hype from Anthropic’s PR team. It was a local, specific, methodologically tight piece of evidence that a general-purpose AI tool, used by a non-specialist, can now perform at the level of a second-year law student studying full time.

The other number I left the day with was 78%.

That's the proportion of Victorians with a legal need who get no help at all. The community legal sector panel framed it bluntly: most people don't enter the legal system through a lawyer. They enter through fear, confusion, a letter they don't understand, a question they can't ask anyone. Legal need is existential. It sits underneath housing, family, health, financial security. And the response, for most people, is nothing.

If you hold those two numbers next to each other, Claude scoring in the honours band in 8.5 minutes, and 78% of people getting nothing, the question that has dominated the legal AI conversation for the last two years starts to look like the wrong one.

The right question came from Jacob Varghese, the CEO of Maurice Blackburn, on the first panel of the day. His point was simple and structurally important: lower cost models will increase demand. The legal market the profession is anxious about losing is not the market that's at stake. The market that's at stake is the much larger one that has been priced out, locked out, or never showed up. If you can deliver good legal help at one tenth of the current price, you don't shrink the profession. You expand the addressable problem.

Fiona McLeay, the Legal Services Commissioner, picked up Jacob's framing directly in the Q&A at the end of the day.

 "I absolutely agree with Jacob that lower cost models will increase demand," she said. "There's a massive opportunity for the legal profession in this area."

This was the framing I heard repeated across every panel today, in different language. The community legal sector called it a generational opportunity. The vendor panel called it lateral disruption. Kate Fazio gave it its sharpest framing.

 "That looks exactly like the taxi industry," she said. "A profession that is guild-like, that is protectionist in places, and that is ripe for disruption."

She pointed at the UK service Garfield, which delivers AI-only legal work on unpaid invoices for around $100 a matter, end to end, including court lodgment. "That market is gone."

Ashleigh Whittaker from Harvey AI offered the sentence I keep returning to. "The wrong question is whether AI can do it. It's whether we should let it." That's not a capability problem. It's a design problem. It's the work in front of us.

I think this is the conversation the profession needs to be having, and it's not the one most of the profession is having.

The most honest moment of the day came from Her Honour Judge Liberty Sanger OAM, the State Coroner, talking about Smart Brief. It's a pilot her court is running with pro bono support from AWS to generate summaries of coronial briefs.

"We haven't seen an uplift in productivity."

She said it plainly and the reason was straightforward: verifying every statement in the AI-generated summary takes about as long as preparing the summary in the first place. The technology works. The output is usable. But the human work that the summary depends on, the checking, the second read, the accountability, doesn't compress, because it can't.

She was not down on the pilot. She was clear it had been a success on the dimensions that actually matter inside a court: getting her team familiar with the tools, building a sense of confidence and curiosity, creating a shared language for what AI is and isn't useful for. She talked about it as cultural infrastructure, not labour-saving infrastructure.

I've been making versions of this argument inside Portable for a year, and it was useful to hear it said publicly, on the record, by a sitting judicial officer running a real pilot and by a practitioner who's coached hundreds of lawyers through this transition. The productivity story is the wrong one. The capability story is part of it. The wellbeing story is the bit that's been missing.

What an organisation gains from putting these tools into the hands of its people isn't a 20% time saving. It's a workforce that understands what it's working with, and a culture that knows the difference between the parts of its work that can be assisted and the parts that can't, and a profession that has slightly more of its emotional bandwidth available for the human side of the work.

That's harder to put in a business case. It also happens to be the only honest version of the case.

The day ended with Fiona McLeay, the Victorian Legal Services Board and Commissioner, taking the audience mic off the cuff and laying out her perspective as the regulator. She's in the final 18 months of her second term. She said she was going to be frank. She was.

She opened with a direct rebuke of the framing she'd heard from earlier speakers.

 "I was slightly concerned to hear one of the last two or three people say one of the great things about using AI is we can do even more work for our clients. That's not the leading approach I would encourage you to take. The levels of wellbeing in the profession are among the worst of any profession in the country, and it shouldn't be that way. Doing more work using AI is not going to help that."

This is the regulator telling practice leaders to design for less work, not more. It's worth sitting with. She spoke about the landscape for AI and innovation in Victoria. The Board has deployed around $5 million in grant funding to date on generative AI and justice technology. Justice Connect, the Office of Public Prosecutions, the Coroners Court, Fitzroy Legal Service, Victoria Legal Aid, the InTouch translation tool for migrant women experiencing family violence. The FY24 program was $600,000. The current FY27 pipeline being assessed is $12 million. That's a twentyfold year-on-year ramp. On top of that sits an impact investing pool of around $100 million, with some components earmarked for justice tech.

A regulator is not supposed to also be the largest single funder of innovation in its sector. But here we are. And she was clear that the work she's funding is the work she thinks the profession should be doing.

She spoke about the Legal Profession Uniform Law and the broader regulatory framework that governs who can practice, what counts as legal work, and how legal services are delivered in Australia. 

"It's nineteenth-century regulation, in twentieth-century legislation, which doesn't work in the twenty-first century."

She called the framework "cumbersome, complex, inflexible, and increasingly irrelevant." Her position, made plain, is that it draws lines that no longer track the work. Asked what she'd change with a magic wand, she said this: "I think we need a regulatory framework that regulates the market for legal services.

"I don't care whether it's coming from a robot or a person. What is the actual point of the information? Is it providing people with legal help?"

That is the design brief for the next decade of justice services in this country, spoken from the office that holds the pen.

A few things I'm sitting with after the event as I put this together. 

The first is that the AI conversation in legal services has finally moved past the demos. Today wasn't a vendor showcase. The most useful moments were a sitting silk admitting Claude passed his exam, a State Coroner admitting her pilot didn't deliver the productivity story she'd been pitched, the CEO of a major plaintiff firm naming demand expansion as the actual prize, and a Commissioner admitting the regulatory frame she works inside is the wrong shape for the work in front of her. That kind of honesty changes what's possible in the next round of conversations.

The second is that the design question, the one Portable has been working on for fifteen years across justice, health, government and social policy, is now the right question for legal services. Not what can the AI do, but what should we let it do, for whom, with what trust architecture around it. Ashleigh Whittaker's "should we let it" sentence is the design brief. The capability is solved. The design is where the next decade of work is.

The third is the gap between the people in the room and the people not in the room. There were a hundred lawyers, regulators, vendors, government officers and judicial staff at the LIV today. They were almost all from organisations that have the resources and the latitude to be at an event like this. The 78%, the Victorians who get no help at all, were the subject of every panel and weren't in any of them. The work that follows from a day like today has to be designed with them, not designed about them.

Portable has been working in this space since 2012. We helped build amica, the AI-supported separation tool used by tens of thousands of Australians. We've built case management tools, navigation tools, intake tools, and policy research in justice. What changed today is that the people holding the levers in Victorian legal services, the regulator, the courts, the bar, the community sector, the academy, were in the same room, saying compatible versions of the same thing.

Eight and a half minutes is the headline. Seventy-eight per cent is the brief. Twelve million dollars is the budget. Wellbeing is the design constraint. And the regulator is asking us to design what comes next.

I'm in.

Andrew Apostola is CEO and co-founder of Portable, a Melbourne-based design and technology consultancy that works with government, justice, health, and social policy organisations. Portable has been working on the design of justice services since 2005.

This article was written with the assistance of AI.