Legal AI
Ask anyone what they pay a lawyer for and you will get the wrong answer. Most people will say the contract, the opinion, the letter. The paperwork. But the paperwork was never the point.
Artificial intelligence now produces paperwork faster, cheaper and in cleaner English than most professionals. It drafts contracts in seconds, summarises a thousand pages before your coffee goes cold. Which raises a question my profession would rather you didn’t ask. What, exactly, is left to pay for? Quite a lot, as it happens. But not what most people think.
Rarely does anyone walk into a lawyer's office solely with a legal problem. They walk in with a partner they no longer trust, a family business and a son who is not ready, a deal they want too much, a letter from a regulator that frightened them. The legal question is merely the shape the problem has taken by the time it reaches the desk.
Type that question into a machine and you will get a competent answer to the wrong question. The adviser's first job is not to answer. It is to find out what is actually being asked. And most of the time there is no right answer to find. There is the position that wins the case and loses the customer. The clause that protects you perfectly and kills the deal. The claim you are entitled to bring against your own brother. The machine will tell you what the law allows. It cannot tell you what you can live with. That part is the only thing on the invoice that was ever worth the money.
Then someone has to put their name at the bottom and be answerable if it goes wrong. The court will not accept an apology from a chatbot. Professional liability insurance covers a person, not a program. And a signature, unlike a machine, works backwards. Knowing your name goes at the bottom changes every line above it.
And even if judgement could somehow be outsourced, there remains the awkward problem of accuracy. A Stanford study found the leading legal AI research products still inventing authority in roughly one answer out of six. They cite cases that do not exist. Courts in England and America have already sanctioned lawyers who filed machine fabricated cases without checking them. The judges were clear about where the fault lay. Not with the software. With the human who signed.
There is one more thing worth saying, and it is not that the machine cannot say no. It can, if you think to ask. That is precisely the catch. The client who most needs to hear no is the one asking how, never whether. A good adviser interrupts, says no to a question that was never asked, and knows it may cost the fee and sometimes the client. The machine waits politely inside whatever question it is given. Ask it whether, and it will answer. It will never ask the question for you.
Advisers of that kind are made, not born. Judgement comes from doing the work badly under supervision and being told why. Every senior lawyer, doctor, architect and accountant you trust today was formed that way, through hundreds of hours of mistakes, corrections and the occasional quiet humiliation. Now the machine writes the young professional's first draft, and it looks good. Too good. The supervisor finds less to correct, so teaches less. The muscle never develops. The work looks fine. The person is not. Ten years from now, that is the adviser sitting across the table from your children.
The answer is not to keep young people away from these tools, which is impossible, but to turn the tools against themselves. Give the trainee the machine’s draft and ask them to tear it apart. Find the case that does not support the point. Find the exception the software smoothed over. Explain why it is wrong. It is harder than writing from scratch. It is also the job. They will spend their working lives checking machines, not competing with them.
In a country this size, all of this is personal. We do not choose advisers from a directory. We know them. We know their fathers. Trust here is not a slogan, it is the entire market. Which is why the question that matters over the next few years is not whether your lawyer, your accountant or your notary uses artificial intelligence. They all will. The question is whether there is still a person behind the page who read it, understood it, and is willing to answer for it.
The machine will keep getting better at producing the work. It will never answer for it, because answering needs something at stake. A name. A licence. A reputation in a country this size. The machine has none of these. The right adviser has all three. It was never the paperwork.
Article by Dr Ian Gauci
This article was first published in The Times of Malta of the 21 June 2026.
Photo credits: Times of Malta
