Peter Littlefair

As a smaller legal jurisdiction, Scotland is often behind the United States and England in its take-up of new tech. AI is no different and a range of options are rolled out into firms already. However, I suspect it can be more difficult to justify a significant spend for smaller firms on bespoke systems.

One area where AI could have an impact in Scotland is handling data internally and improving reporting measures. These sorts of administrative tasks are an obvious target.

In relation to civil litigation and case management, there will likely be improvement of bundling software – programmes that collate all relevant papers for a case and ‘bundle’ them together in a coherent, searchable form. A large language model can then search for other relevant cases, trends, and perhaps even predict the likelihood of your case succeeding. We know of platforms adopting this approach.

Case prediction has been seen in Scotland before with Colossus, a model used by insurers to predict compensation payments against litigation risk to try to control settlements. In a time where inflation is high, there may be significant benefit in this approach for insurers.

There are a number of stages with AI at which you need to be cautious. Firstly, it works best when asked the right questions, phrased in the best way, so-called ‘prompt engineering’. Yet even if you ask the best question, the AI programme might still not give the best answer. There are a number of articles about ‘hallucination’ – creating an answer even if inaccurate.

At every level, lawyers will need to review automated advice. Who would stand up in court having not read the relevant papers and risk being asked a question they can’t answer? It’s not really the done thing to say ‘I’m not sure. Can I ask the bot?’

Predictive modelling will inevitably focus on larger jurisdictions, so will all relevant Scottish cases be included? Or will AI throw up cases from elsewhere which might be of interest but not persuasive in Scottish courts? In my view, predictive AI models are interesting, but not good enough yet to make inroads into civil litigation. ‘Yet’ is the important word.

I use a travel analogy. For city breaks with budget airlines, most people book online. For a luxury holiday of a lifetime, you’ll use a travel agent. In the law, for simple work you might choose AI but for complex and challenging cases I’d recommend relying on human expertise.

Inevitably, AI will improve and be able to fulfil more legal functions effectively. I’m not sure all firms are ready for that. And I don’t think the popular narrative – that AI does the boring stuff and frees up human creativity to increase profitability – is necessarily true.

As lawyers, our business value is the time we spend on client work. If AI does become effective you still need to validate and interpret it, but AI will reduce time spent on cases. What does that mean for what you charge, and for legal business models?

We have to think about that, not bury our heads in the sand. But we also need to remain sceptical, and questioning, of what AI can (and can’t) do, to protect our often envied legal system.​​​​​​​​​​​​​​​​​​​​​​​​​​​

​Peter Littlefair is a senior litigation associate at Balfour and Manson. This article first appeared in The Scotsman.