Companies and law firms have spent decades trying to get computers to do legal work with decreasing levels of human intervention, investing huge amounts of time and money in a problem nobody was sure could ever really be solved.

Then in autumn 2022 ChatGPT was released and the world changed – all of a sudden anyone who could type could access the seemingly near-human power of artificial intelligence. Some experts knew it was coming but for the rest of us it was a sudden and unsettling change that made us ask some pretty searching questions about the nature of our work.

For the legal function of large organisations it presents an opportunity to expand capability, to focus precious human resource on difficult, high value problems and to realise previously-hidden commercial opportunity.

But it also demands we answer some fairly big questions – about ethics, intellectual property rights, the systematisation of human bias and what the legal and moral limits of automation should be.

These are old questions, but things are moving fast.

Alastair Morrison, who is in charge of client strategy at Pinsent Masons, said that AI has the potential to expand the capabilities of forward-thinking legal functions.

Morrison Alastair

Alastair Morrison

Head of Client Strategy

Generative AI is augmenting the capabilities of the legal team, in contrast to becoming legal team automation

“The universe of law is expanding at an ever-increasing rate,” he said. “Regulation is not going to decrease - what industry isn’t regulated? With that comes increased compliance and increased investigation. This technology offers the opportunity to get your arms around things you otherwise wouldn’t get your arms around, such as getting a better risk management process in place. Generative AI is augmenting the capabilities of the legal team, not replacing them and not automating things, but being a very powerful tool to get stuff done that wouldn’t otherwise be achievable.”

“For in-house legal teams we often talk about a polycrisis or a permacrisis where the sheer range of issues being dealt with means they don’t have the bandwidth to deal with it all. You put stuff away that you would love to deal with but you don’t have the time, the resources or the finances to address.

I can see applications for this for projects that you don’t have the time or resource to do, like putting improved risk management processes in place that give you institutional integrity and resilience,” said Morrison.

AI before GPT

AI systems can already do a lot in the legal field, such as predict trial outcomes and judges’ rulings extremely accurately. According to Lucy Shurwood of Pinsent Masons, who’s been working on using technology to deliver legal services for over a decade, systems can now extract precise information from contracts quickly and easily.

“It is making it easier to extract the relevant information from a contract so that it can be reviewed and evaluated. This works for anything to do with contractual terms. The key terms that a client might be interested are ‘when can I terminate the agreement?’; ‘what's the governing law of the agreement?’; ‘if I want to transfer the contract, can I do that? Are there any conditions?’. It generally arises in an M&A transaction or is driven by regulatory change that a company has to comply with. That’s when they need to understand what the contractual position is in a very large number of contracts. So it's about being able to extract information really efficiently and support decision making,” she said.

Know your terms

All the different names used for AI can be confusing. Here’s a quick guide:

 

Discriminative AI

predicts groupings of data and can be used to classify types of information, for example cats and dogs.

Generative AI

predicts how data is spread across groups and can produce new examples within a context.

Deep Learning

a type of AI based on neural networks that maps inputs and outputs through multiple layers of numerical weights that have been pre-trained on example data.

Language model

generative model that scores probabilities of words across a sequence. Predicts the next words in the sequence.

Large language model

a language model with massively increased scale using deep learning and trained on massive amounts of text.

 

Shurwood said that generative AI makes this much faster, when enables better, quicker decision making and saves a lot of time and effort that can be expended on other tasks.

“In the old days the way this would have been done is take a pile of 200 contracts and have a lawyer read them all and produce a written report – you then have 200 reports, and a client doesn’t want to read 200 reports. So someone would have to read them all and write a summary, which could take days. We now generate all of that automatically, so all of the information that's extracted from the contracts combined with where relevant the lawyers input is all stored in a database and literally at the push of a button we can create a summary report for a client.

“For example I have reviewed 250 contracts to find out whether or not they can be transferred, and I've done that in four minutes in one case,” said Shurwood. “So it's much more cost effective, it's much faster, but it's also giving the client much more relevant information.”

Everything changes – enter GPT

But something fundamental has changed in the nature and power of these systems, said Orlando Conetta, who is in charge of product development at Pinsent Masons. “We are witnessing the rise of remarkable capabilities in language processing, all emerging from the same basic operation – predicting the most likely word to follow from a previous set of words.” he said. 

“This leap in performance was made possible due to the groundbreaking ‘transformer’ architecture, first developed in 2017 by researchers at Google and in academia. It has dominated deep learning research since then, and it was this approach that was used at massive scale by OpenAI to create the ChatGPT platform that astounded the world with the breadth of functions it could reliably perform,” said Conetta.

“OpenAI’s GPT - Generative Pre-trained Transformer - model is so powerful that it has changed how language processing is done. In the past, new models would require extensive training on large sets of data oriented towards the task one was trying to perform such as sentiment analysis or clause tagging,” said Conetta.

 

Hear more about AI and the legal function on the Brain Food For General Counsel podcast.


Previous systems needed extensive and specific training in the exact functions you wanted them to perform and environments and types of information you want them to perform in. That has changed.

“Today, users and engineers alike can use GPT directly for their tasks without further training of the system. They do this through the new discipline of prompt engineering - setting out the instruction, context, input data and output format that best serves the task. These can even be combined in sophisticated chains or integrated with external data sources,” he said. “Training is still an option though, and you can fine tune a model further, making it more tailored to your needs, building on the power of the underlying model with a fraction of the training data previously required.”

These are the changes that are causing legal leaders to wonder whether the fundamentals of how legal functions operate are going to change.

Shurwood said that the impact has been immediate. “We're seeing from the early experiments with generative AI that it can take things that were either very, very difficult or very, very time consuming or both and made them much more accessible. So for example, we have been working on extracting a particular piece of data from contracts for about two years. We've been trying to improve the models, make it more reliable and it has always been very, very challenging to get that particular data point out. And it now looks as though generative AI will be able to do that with maybe a couple of weeks tuning. So it's the difference between investing an awful lot of time and expense in being able to do something versus being able to fine tune a system to do it relatively quickly and reliably.”

That kind of power has the potential to change something as fundamental as how contracts work, said Morrison. “It is interesting to see the evolution of contracts. They used to be stuck in a drawer and you would only look at it if you had a problem. They evolved to being much more of a front end management tool – if you think there is a problem emerging give me notice so we can work out the cost and see what’s going on.”

I have reviewed 250 contracts to find out whether or not they can be transferred, and I've done that in four minutes.

“If systems now start talking to each other then contracts can talk to programme management tools which can link into reports of live events on a project, such as an adverse weather condition which might entitle people to take longer to complete the job. There are sometimes disputes about that. But you could see generative AI instantly resolving that and it being wired into a programme management tool which would change automatically,” he said. “So you might see far fewer disputes emerging – the impact of change and variation can instantly be adjudicated and resolved.”

Normally a dispute arises because people are in denial about how a job is going and they’re wishing for the best. I don’t think in this case these even become disputes – it’s like micro or nano-resolution. It’s not even a dispute, it’s a factual circumstance occurs and the consequence is impact on the time and programming and the consequences are instantly dealt with and there’s nothing to be in dispute over. That’s far down the line but it is interesting, and applications that we can’t envisage at the moment will emerge,” he said.

A tool for transformation

Many legal functions are already working on digital transformation of how they operate according to University of Miami law professor Michele DeStefano, who regularly interviews general counsel about trends in legal operations. So there is already a process that AI can fit into.

“The biggest trend for in-house legal departments right now is digital transformation, and what in-house legal departments are trying to do is create a way to manage all of the intake, all of the requests from the business,” she said. “One of the most common complaints I hear when I interview general counsels is that they get the same requests by the business yet there is no streamlined way to provide the answer. They need to figure out a way to use digital tools so that they can free themselves up from some of the work on easy answers that could be automated so that they can get to the strategic work.”

“The GCs that are starting to digitally transform are thinking about ‘how am I going to create a front door to legal?’ meaning you as a business person has one place or platform that will help you with your problem through chat or an AI tool so you get faster results,” said DeStefano. “The GCs that are with it right now are restructuring their departments, hiring legal ops people, looking at alternative legal service providers that can use technology to pull everything together so you could mine your contracts for provisions that can save the company a lot of money. And to find secondary opportunities – maybe there’s something in the contract that gave you rights to something else that no-one saw.”

Morrison Alastair

Alastair Morrison

Head of Client Strategy

More regulation goes into buying a cup of coffee than goes into generative AI, so we have to get our arms around that

The general public was stunned by the capabilities of ChatGPT3 in autumn 2022. Reaction of the technical community to ChatGPT4, released in March 2023, has been one of astonishment. Progress is fast and will get faster as AI systems are improved not just by human engineers but by AI systems themselves.

The initial challenges for legal teams will be practical. Because systems are ‘trained’ on existing knowledge, organisations will have to take steps to make sure their use of AI doesn’t breach other people’s intellectual property rights. Unintended consequences proliferate, so getting requests and prompts right is essential. And some regulation and control is needed, as is the case with other powerful and potentially destructive consequences.

But the much bigger questions for organisations are about how this technology will change the very nature of their work. Conetta is certain that lawyers will still be essential to any legal process. And DeStefano said that the value of these systems isn’t as a search engine, but as an ideas generator.

And some control and regulation will be essential.

“Before we get too far ahead of ourselves we need to think about the ethical implications of generative AI, there needs to be a lot more thought about this than there is – somebody said to me that there’s more regulation goes into buying a cup of coffee than goes into generative AI, so we have to get our arms around that,” said Morrison.

We are processing your request. \n Thank you for your patience. An error occurred. This could be due to inactivity on the page - please try again.