AI in the Legal Practice: from Hype to Reassessment

Whereas AI still seemed mostly experimental a few years ago, it is now a technology that legal counsel can no longer ignore. AI systems are constantly improving their skills in text processing, generating draft documents and supporting research. Moreover, AI is expected to keep developing in the coming years and become able to take over more and more tasks currently still performed by humans. Those who do not learn to work with AI now risk missing the boat. Although it is always hard to assess a technology in progress, we can already draw some tentative conclusions from our own experiences with AI.

We start by stating the obvious: AI is very handy indeed. It may be worthwhile to follow an AI course or read about it, but its real possibilities are only discovered through personal experimenting. Not all experiments have to be in a business context at all. It is even easier to recognize AI applications in your personal life. Plan a trip, prepare dishes based on the contents of your fridge, read menus in a foreign language, etc. Once you know and use the options, the application at work becomes a lot simpler and more logical.

Here are a few easily accessible ways for lawyers to start with in their work:

• Perform market research into your client or opposing party. Ask an AI system to profile a sector or a company. You will quickly get a first impression of their position, reputation or legal sensitivities. Of course you have to check the facts, but this can be a good starting point.
• Research the market of your opposing party. We recently defended a client in a lawsuit in which it was alleged that the client's general terms and conditions were uncommon and unreasonable. With the help of AI, within a few hours we managed to compile an overview of the terms and conditions that competitors around the world use, in various languages. This way, we could prove that the sector uses similar terms and conditions throughout the world. Without AI, this would not have been possible, or at least not as quickly and cheaply.
• Let AI trace inconsistencies in an argument. Take written pleadings of your opposing party or a draft text of your own and ask the AI where the reasoning is weak or inconsistent. This may expose blind spots remarkably fast. It is a rather safe application, as it requires no deep legal knowledge. Knowledge of language suffices and this is a thing that AI systems are already very good at now.
• Let AI help you write publications. Talk to an AI about the article you want to write. Who is it intended for, what is the style, how long should it be, which topics do you want to discuss? Let AI write a first draft and keep working on it together until it is good enough to complete it yourself. 

• Let AI improve your texts. AI is already pretty good at this. You can use it to improve consistency in longer texts, turn passive into active, and clarify sentences. For emails, you can ask it tailor the message to the recipient and even check for cultural differences. Not everyone will appreciate Dutch directness.

   

For those who want to go one step further, there is much more to do:

• Experiment with your own legal “agent”. Compile a distinct data set – e.g. a series of judgments, policy documents, or internal memorandums – and make the AI generate answers from this set only. You are now imitating a digital junior associate who quickly gives you the correct passages. This will teach you how context restrictions work and which obstacles the technology encounters in practice.
• Train an AI in your own style. Especially if you use Copilot and allow it access to your Outlook and documents, you can use AI to use a tone of voice that suits your own advice, contracts and procedural documents. You will have to put in some work here, but it is definitely worth the investment. It makes the input of AI much more usable and recognizable, which can render your own work more efficient.

The more you use it, the more you will get an idea of where AI is truly helpful, and not just a new shiny toy. You will also learn to recognize signs of a ‘hallucination’: an answer that sounds convincing, but is factually incorrect. You will experience the importance of good prompting. Perhaps even more importantly, you will practice what you and your peers preach. Not only can you explain what AI can and cannot do, you can also demonstrate how you use AI in practice and where its risks lie. 

Despite its potential benefits, in our experience AI should not be used to perform strategic or really substantive legal work, since the quality of its output is too uncertain. Hallucinations are still common and can only be recognized by those with an excellent command of the subject matter. For example, when we asked an AI to list the main lessons from a judgment, it came back with answers that were largely correct, but also introduced a completely new consideration or a wrong reference to existing passages. This is very risky for lawyers. It is easy to spot mistakes when 75% is obviously wrong, but a lot harder when it is only 2%. AI systems also tend to produce near-misses of the core of a question. They cannot replace years of experience and the ability to weigh interests, strategy and nuance. The human factor remains crucial, which should be a comforting thought for lawyers. At least for now. 

At the same time, it is clear that AI will only get better. Legal departments and law firms will therefore have to ask themselves where their added value lies. If an AI tool for analysing contracts or understanding regulations can already get a business or client a long way, what will a lawyer still be able to add? And is the added value of certainty or added nuance worth the effort and money for a client?

Incidentally, this is not an entirely new phenomenon; the arrival of reliable search engines too extended the accessibility of legal knowledge. This may have changed the nature of the work of lawyers, but it has not diminished the demand for legal services. Quite the contrary, in fact.

This is why our expectations are positive. We believe that AI can further improve the services of lawyers and allow them to do more meaningful work within the time or the budget available. They can present themselves as the senior advisor who performs the last check and makes adjustments at the start or the end, or who provides better services in less time. We also see opportunities for tech-savvy lawyers, who can improve their services so much with the help of AI that their clients or business are left with much less work to do in that respect. This gives back control to the lawyers, which also contributes to risk control. However, this implies a clear challenge to the lawyers: those who pick up the gauntlet now can make the difference later.

In the years to come we will undoubtedly witness further steps, not only in technological progress, but also in the developing legal frameworks. The recent entry into force of the AI Regulation has made it clear that legislator and supervisor alike appreciate the importance of reliable and transparent AI. This is a double challenge and opportunity for the legal sector: learning to work with AI and ensuring that the deployment of AI as such meets the new requirements. Interesting times are ahead.