Anything you can do, AI can do better

We now create information so quickly that no human being could possibly hope to keep up with absorbing it all.

This creates a fresh set of challenges for sectors like law, which rely heavily on subject area knowledge, expertise, and constant learning to stay current with developments.

But if technology has caused this problem, it also presents a possible solution: artificial intelligence (AI).

AI can apply machine learning and cognitive computing techniques to understand the unique context of law and the relationships that a firm has with its clients.

Huge advances in mathematics, algorithms, and raw computing power have changed the game. Until recently, it was impossible for computers to analyse unstructured data at scale or with speed.

Paradigm shift

Now, the implications for the legal profession are profound. “The technology can connect dots and analyse a scenario more holistically than ever to see patterns and trends that no one has seen before,” says Brian Kuhn, co-founder of IBM Watson Legal.

Kuhn was speaking at a briefing in the Law Society in September. The technology company Sure-Skills held the event with guest speakers from its partners IBM and Filament AI to share knowledge about AI and explore practical uses for legal professionals.

AI application

A regular speaker at conferences worldwide, Kuhn is well placed to comment on legal and technology trends. He invented the AI application, IBM’s Outside Counsel Insights, specifically for legal professionals, and he has held more than 140 workshops to explore how AI can add value to a client’s practice.

“The number one use case in the legal domain is knowledge management – leveraging what a lawyer has done for clients in the past, in the context of the issue at hand,” he told the audience in Dublin.

Unstructured data

The legal sector, like many others, may soon find AI is a necessary – not just nice to have. Industry estimates say the amount of data is doubling every 12 hours. Kuhn says that AI differs from traditional technology because it can analyse unstructured data that, until recently, has been invisible to machines.

He estimates that around 80% of information is in the form of unstructured data.

Computers have traditionally needed information that’s organised and can be put into a database. ‘Unstructured’ data, such as written text, is much harder to understand. But think of the possibilities of technology that can ‘read’ and comprehend text in legislation or in contracts.

Applying AI in a legal practice

Applying AI in a legal practice would mean that a company could understand, from comprehending precedents and case law, whether its actions would expose it to litigation. “With more data come more patterns; signals emerge from the noise,” he says.

However, using artificial intelligence isn’t as simple as flicking a switch. It won’t work straight out of the box: “The way you purchase AI will be different to how you purchase software, where it is one size fits all,” says Kuhn.

“Now, vendors want to understand your needs. The benefit is that AI is now capable of taking those needs and creating relevant AI models. That makes it much more likely to succeed.”

And therein lies the rub: although it’s a boom time for technology in the legal sector, with thousands of potential products to use, there’s a snag: “Most AI solutions are not great because they don’t incorporate legal knowhow. Most of them are not trained by lawyers.”

He points out that human language can be ambiguous; its meaning is highly dependent on the context of a culture or an organisation. “A machine doesn’t understand if you say: ‘I’m feeling blue because it’s raining cats and dogs outside’,” he points out.

Recognising patterns

For this reason, the second part of the seminar was dedicated to looking at how to ‘train’ AI to understand and recognise patterns in data. Andy Feltham, a former master inventor at IBM and now leader of specialist consultancy Filament AI, says the key to building useful AI is not to think about it exclusively as a technology problem to solve. “It’s about understanding your business, not just data science. You can’t just read a book and do AI. There are good and bad ways to do it,” he cautions.

“Unless you apply context to the theory, you get the wrong result. You need to apply your expertise and understanding to make this system work. AI is not intelligence: it has the same relationship to intelligence as artificial flowers have to real flowers. The machine can find the patterns in data that we wouldn’t otherwise be able to,” Feltham says.

The ‘teaching’ process

This ‘teaching’ process involves giving examples to the AI tool, such as going through a process of annotating data. This effectively means classifying information so the system understands what it means and can learn to identify patterns.

“AI is different to developing an application. It’s about iterations, sharing, publishing and monitoring performance. Then you observe when it goes wrong and add that knowledge back in. This is about building models to encapsulate what you’ve done in a repeatable, scalable way,” says Feltham.

Relevant data

That’s why it’s important to feed the AI tool with data that is relevant to the profession: that might be information about prior cases that a firm already has. Feltham made an analogy with the music service Spotify. It knows characteristics about all the music in its library and uses them to suggest other bands based on patterns in someone’s listening habits. Those recommendations are unique to each listener.

In a legal context, that knowledge would result in a digital assistant that understands natural language queries and would search through a firm’s knowledge base of previous cases. “In a legal space, that might be typing ‘show me the documents that date from this time’. Being able to describe the report you want could save hours of time with Excel,” says Feltham.

Fearmongering

The difficulty with discussing AI is that it tends to have hype or fearmongering for company. Both speakers at the seminar played down the direct risk to lawyers’ jobs. Instead, they said it provides opportunities to do more for their clients.

Feltham claims that saving time on research will liberate lawyers to meet their clients more often. Or, it will give them a closer understanding of a client’s problem because the AI can ‘read’ and analyse more material than a human researcher can by themselves.

Extra work from clients

Kuhn thinks that AI could help legal practices expand, not contract as some fear. “Law firms that are considered as innovative are 66% more likely to get extra work from clients. Law firms will be able to offer clients far superior legal services. With cost-effective delivery and repeatable legal services, you would be able to differentiate against many different firms,” he says.

AI is already delivering real value and saving considerable money in other sectors. “This flux is coming to this industry,” Kuhn adds.

“More people think it’s positive. It forces delivery of excellence; it forces the client and provider to have conversations about how the technology will affect your relationship and how you can be part of that.”