The most important AI skill for junior lawyers is knowing when to disagree

Reading time: five minutes

Earlier this summer, an article in the Guardian asked who should be responsible when doctors rely on AI tools that contribute to mistakes. Although the article focused on medicine, it’s a much broader question that also applies to law: when a professional relies on AI, where does assistance end and professional judgement begin?

I graduated in 2024, when generative AI tools like ChatGPT were only just becoming part of the conversation; people were wary of them, and my university strongly advised against using AI to complete assignments. Two years later, they’ve become an everyday part of life.  

I can understand why: used appropriately, AI can save an enormous amount of time. We can use it to work through information, identify issues that need to be investigated and even stress-test an argument.

But that efficiency can’t replace judgement or experience.

Interested in finding out more about AI in the legal profession? Register now for our masterclass with Baker McKenzie!

AI provides support, not an answer

In some ways, AI is like an X-ray machine; it can provide valuable information, but it’s only one part of the picture. The doctor still needs to examine the patient and consider their history to suggest the best course of action.

In the same way, an AI tool can help us work through a long document or spot points worth investigating. Remember though, it doesn’t know the client; it can’t understand their commercial priorities or see every nuance.

More importantly, it can’t take professional responsibility for the advice you put your name to.

AI tools aren’t bound by the same duties and professional standards we are. I can’t defend a legal argument by saying that AI suggested it; I need to understand the argument, verify it and justify my position.

The court’s judgement was clear: it’s still the lawyer’s job to check. In fact, a ruling from the Sheriffdom of Tayside, Central and Fife at Kirkcaldy earlier this year indicated that submitting AI-generated content could lead to being held in contempt of court.

That’s why it’s so important to verify everything. But incorrect information isn’t the only risk to using AI tools. The bigger risk, and the harder one to overcome, is that the output can be factually correct but still not right.

Verification is just the start

One of our partners recently told me a story about using AI to review a contract. The tool hadn’t hallucinated a clause or fabricated anything; it just produced an interpretation. Because of his experience and legal knowledge, he was confident in his own interpretation and was able to put the AI’s ‘view’ aside.

I’ve also experienced for myself instances where AI has cited legislation from a different jurisdiction (despite being prompted otherwise) or provided a source that doesn’t include the information it claims to have.

Thankfully, I’ve always been taught not to take what AI says at face value. As a trainee, though, it can be easy to assume that the polished AI interpretation must be right.

You can’t rely on AI to reach a conclusion. It is, however, a great resource for speeding up the process of reviewing lengthy contracts or complex trust deeds: I prompt it to identify information based on key words and provisions, which I can then verify myself. I also always skim the rest of the document to ensure nothing was missed.

And verification is just the start: a source or quotation can be accurate, but the AI’s analysis might be incomplete or miss an important piece of context about the client’s circumstances.

Should you use AI in law firm applications? The Oracle answers.

Form your view before asking AI

Trainees and junior solicitors need to become comfortable with technology, but we also need to develop the underlying skills to supervise it.

If we reach for AI before forming our own interpretation, we miss the chance to develop our own legal judgement. A better approach might be to draw your own conclusion, then test it with AI. Investigate any differences; it’s a great learning exercise.

As a trainee focusing on cyber and tech law, I should also mention here that responsible use of AI begins long before you actually use it. You should always confirm it’s an approved system and your firm’s policies allow you to use it for this specific task.

Clients pay for judgement, not just information

Learning to use AI appropriately is already part of becoming a capable modern lawyer. Using it well means knowing its limits and having the confidence to disagree.

I know it can be tempting to lean on AI, especially if you have a large caseload and looming deadlines. But clients don’t come to lawyers for information they could generate themselves. They come to us for analysis and our ability to use our understanding of the law and its nuances to create a recommendation.

AI can help us form that recommendation, but can’t do it on our behalf.  After all, it’s not the tool’s name on the paperwork. It’s yours.

Lucy Hanlon is a trainee solicitor at Anderson Strathern LLP.