Automate the preparation. Protect the opinion. Disclose the difference.

Think about everything that goes into producing a medicolegal report. Then ask how many of those hours needed a medical degree. Records assembled and read. A chronology built. Imaging cross-referenced against history. Documents indexed. Formatting, structure and boilerplate. Then the examination, the reasoning, and the opinion. Only the last part needs the doctor.

This week Peter Gibson, President of the ALA in Queensland and Head of Legal Innovation at Shine, put a question to the industry that has been sitting just under the surface for a year. If AI is used in a medicolegal report, what exactly should be disclosed? His framing was the useful part. There's a difference between using AI to organise records and prepare a chronology, using it to help draft a report from opinions the expert has already formed, and using it to generate or influence the opinion itself. Three very different things, currently collapsed into one word.

What's beneath the line

That list from the top of this piece is not a small part of the job. On a complex matter it's most of the elapsed time, and none of it requires clinical judgement. All of it currently consumes the hours of people whose clinical judgement is the scarcest resource in this industry.

Issue 07 covered how experts decide whose work to take. Issue 08 covered what happens to supply when the fee stops matching the value of the hour. Both come back to the same shortage. There is not enough senior specialist time, and a meaningful share of what exists is being spent on work that doesn't need a specialist at all. So the disclosure question and the capacity question turn out to be the same question, looked at from two ends.

What that actually buys

Give an expert those hours back and they don't disappear into a margin. They go somewhere useful. More time in the appointment with the injured person, which is the part that decides whether the report is worth anything. More time on the causation reasoning that gets tested under cross. More capacity in a market that is short of it. This is the case for AI in medicolegal, and it has nothing to do with machines writing opinions. It's about giving the opinion more of the expert's attention, not less.

I should say plainly that I work for a provider who stands to benefit if the industry lands where I've just argued it should. Weigh that however you like. I'd make the same argument if I sold nothing at all, but you're entitled to know I'm not a disinterested party.

The caveat that matters

Greg Spinda made a point worth carrying forward. If a chronology or summary contains errors, those errors can infect the opinion built on top of them. Careful use is still required below the line. He's right, and he added the part people miss. That risk applies whether the chronology was produced by AI or by a person. A bad summary is a bad summary. The question isn't whether a machine touched it. The question is whether anyone checked it.

Australian courts have now dealt with practitioners over AI-generated material that nobody verified before it was filed. Not New York. Not London. Here, under our professional standards. The failure in those cases wasn't the technology. It was the absence of a human check between the output and the filing. Automating the preparation only works if somebody still reads it.

What disclosure should actually say

A disclosure that says AI was used tells the reader almost nothing. It's a box, and boxes get ticked. If the three tiers are meaningfully different, and they are, then the disclosure has to distinguish between them or it isn't doing any work. Something closer to this:

AI was used to assemble and index the records and to prepare the chronology. The opinion, the findings, and the reasoning are the author's own. All AI-assisted material was reviewed by the examining specialist before the report was finalised.

Three sentences. A reader knows what happened, a court knows what to weigh, and the expert has said plainly where their own work begins. Nothing in that is difficult to write, which raises the obvious question about why so few reports carry anything like it.

The framework, one more time

Augmentation. Transparency. Verification. Augmentation is everything beneath the line. Verification is the human check on all of it, because Spinda's caveat is the one that bites. Transparency is the disclosure that tells the reader which is which. Automate the preparation. Protect the opinion. Disclose the difference.

The bottom line

The industry is going to spend the next year arguing about where the line sits. My view is that the line is already obvious and the argument is a distraction from the more useful question, which is what we do with the hours underneath it. Every hour a specialist spends assembling records is an hour not spent with an injured person, not spent on the reasoning, and not spent taking the next matter in a market that badly needs the capacity. That's the prize. Disclosure is what makes it safe to go and get it.

Next issue: the same scheme fee schedules that set what a specialist gets paid also set what a solicitor gets paid. One document. One regulator. Two very different views on what an hour is worth.