UK law firm Weightmans says that the use of Artificial Intelligence (AI) has cut the time needed to review and extract required information from legal documents by 90%.
This means, says the team, that its lawyers can concentrate on directly helping clients, rather than wasting time on admin.
Even better, the system is now fully integrated into the organization’s data warehouse.
This means that useful data from document analysis can be shared across the organization and other applications using APIs.
And according to its Chief Technology Information Officer, Stuart Whittle, that means that the firm is finding new ways to delight its customers.
No client ever said to its law team, ‘Take longer and cost me more.’ You always want to be faster, cheaper, more accurate, and generally better.
Both lawyers and other people in organizations spend a lot of time finding stuff, and we seem to have a way here to make that happen more quickly.
Avoiding ‘prohibitive expense’
With 1,500 staff and a presence in nine cities, Weightmans is one of the top 40 most successful UK law firms.
A full service law firm, it operates in seven business areas - including, insurance, healthcare and private clients.
Whittle says the organization’s journey with legal AI began when, two years back, a major client requested help on a big information discovery challenge.
Specifically, the client needed help to review 1,800 long documents to respond to a regulatory issue.
This problem was compounded by the need to meet a very tight deadline - a mere five business days.
The usual way lawyers would go about this is just to throw people at it - working 18-hour days to go through all the documents. That’s generally a two-stage process, whereby you go through the documents to extract the particular clauses that the regulator was interested in.
That tends to be a paralegal, fairly low-skilled job. And then you pass the output of that to specialist lawyers to review, comment, and exercise some judgement.
The problem: doing this with human labor was prohibitively expensive, as each document was 20 to 40 pages long.
By coincidence, Whittle and his team had just been experimenting with a documentation automation tool that claimed to be able to speed up the first stage of this process - document search and review.
Weightmans offered to run an experiment with this software to try and meet the customer’s tight deadline.
To be honest, the client was very skeptical that a piece of technology could replace human intervention here. But given the lower price point and timescale, we got the go-ahead to try.
The system - now called Litera - did indeed meet the requirements, and the deadline was met.
Subsequently, Weightmans has become more and more convinced that this kind of approach to document search, to support legal work, is effective.
This confidence has been bolstered by significant growth in functionality since that first project.
This initially started as a technology that would just help you extract specific types of clauses from documents. For example, in a commercial contract, there might be a jurisdiction clause that you're interested in, so you tell it that these sets of words tend to be associated with a jurisdiction clause, and it will extract that for you so you check it.
However, the product can now be taught much more abstract concepts - for example, not just finding clauses relevant to a particular clause, but now ‘knowing’ what the jurisdiction and relevant law actually is, in either the US or UK.
That can really speed up my work as a lawyer, but it’s also getting better and better at picking up useful facts and instances from the documents it’s reading.
It could already identify what the rent review clause says, but what I'm interested in may not be the clause itself, but the actual rent amount in the case. Now, it can give me that.
Confidence is key
Whittle and the lawyers he supports would be interested in a data item like rent, because it can often be at the center of a case.
As part of an M&A deal, for example, the company has a significant property portfolio, so looking at leases could be a very material aspect of the final bid price.
Another example is in plot sales, which can involve hundreds of individual sales.
Historically, the client sent over multiple PDFs of such plots, which all contain information that needs to be extracted. Weightmans then created jobs on its internal case management system to manage the work.
Prior to AI, because of all the data that needed importing off the PDF, that could take at least 60-65 minutes to do.
To speed things up, Whittle trained the new system on examples of these types of instructions to extract the needed data and then built an RPA solution that can quickly pull out everything that’s needed.
The PDF is then set aside in a folder monitored by another system that sends it to the AI, with findings checked by a human for safety.
This process alone has been reduced in average elapsed time by 55%, he says - down to 35 minutes.
Another example of why it’s useful to have an automated way to hunt through large volumes of documentation is an employment dispute, where employment contracts need to be looked at in detail.
The use case here is definitely around being able to analyze big volumes of documents quickly and with the minimum of human effort involved - but being confident the system’s always going to give you accurate results.
Whittle has such confidence, as audits consistently show an accuracy rate via this process of 98%.
I would venture to suggest that if we’d had people working long hours on this and getting tired, we wouldn't always hit 98% accuracy doing it manually.
A range of organizational and client benefits
There is obviously a cost-saving element to being able to do this and avoiding use of people, but overall productivity and even scope for better service is raised.
Going through large volumes of documents and extracting clauses is dull as ditch water. And because it's dull and boring, people can lose concentration and miss things and make mistakes. Yes, people still must go through to check the findings, but that process has been much, much, much faster. And so you can handle greater volumes of work.
When things that are repeatable and predictable get done by machines instead, you start to be able to meet historically unmet legal needs, because previously the cost involved was too high using just a people model.
Next steps for AI at Weightmans go well beyond document search.
Whittle wants to explore, for example, potential use cases for large language models (LLMs) in the law - both on the business services side, like speeding up some marketing processes, and on the legal side of the business.