Legal tech teams turn to AI to advance business goals

For a second year, the FT’s Accelerating Business series has looked at the growing contribution of legal services providers in meeting the fast-changing needs of business.

Increasingly, the focus has turned to generative artificial intelligence and its potential to transform legal work. Previous articles in the series have covered the technology’s role in contract review, ensuing data privacy concerns, and those tech providers competing to offer the latest tools to clients.

Last November’s release of the ChatGPT chatbot kick-started months of experimentation to see how generative AI could be applied to legal tasks.

Specialist start-ups such as Harvey and Casetext, which was acquired by Thomson Reuters in August, have built a name for themselves by quickly developing AI services for the legal sector. Harvey struck a high-profile deal with magic circle law firm Allen & Overy in February and another with Big Four accountancy firm PwC in March.

Then, LexisNexis and Thomson Reuters, the established legal research and technology providers, announced extensive generative AI features in May. Both are still trialling their new offerings for a limited number of clients.

However, not everyone jumped in so fast. Legal technology company iManage waited until August to announce it was testing generative AI features, such as a ChatGPT-like “conversational search” function for its document management system, with two of its clients. It added that earlier forms of artificial intelligence — in which it has invested for many years — still work well and can be far more cost-effective for automating routine tasks, such as filing large volumes of emails.

California-based cloud software company Intapp has taken a different approach, meanwhile, and started to integrate generative AI into its marketing and business development offering for legal and professional services firms. The company says it has intentionally taken a measured approach, rolling out small uses for the technology — such as writing email subject lines based on the content of a message — as they become ready, rather than announcing a bigger road map in one go.

And, as the initial buzz surrounding generative AI’s impact on the legal sector begins to subside, a more nuanced understanding of where the tech can be applied is emerging. The case studies below highlight how the legal arms of the Big Four professional services firms, and some from the legal technology sector, are applying the technology to accelerate legal work. These examples show that, when applied appropriately, it can be a powerful tool.

Tom Saunders is a content director at RSGI

Accelerating Business case studies

Over the past six months, the FT’s research partner RSGI, a legal think-tank, has provided examples of how the legal ecosystem is using new technologies and techniques to serve the fast-changing needs of business.

ContractPodAi 
The legal technology business commercially launched its generative AI-powered “legal co-pilot”, called Leah, in August. The tool can help clients review and redraft contracts, and dozens have adopted it since release. Relevant documents are uploaded to the tool and then users can ask questions about the contents of the documents via an AI chatbot. An investment firm client used Leah to conduct a portfolio management exercise, for example, and found that this helped to complete the process in weeks rather than months.

Deloitte Legal
The Big Four professional services firm created a legal generative AI centre of excellence to build tools for internal and client use. Deloitte’s legal management consulting team works with clients on their generative AI digital strategies and on pilots to improve processes such as contracting. One financial services client is trialling generative AI to analyse the difference between the opening and final positions agreed on a contract during a negotiation, to speed up what is often a protracted and costly process.

EY Law
EY has spent $1.4bn over the past three years on building its EY.ai artificial intelligence platform, which combines a proprietary large language model, called EYQ, with other technologies and human expertise to help clients adopt AI. The accounting group’s global legal practice, EY Law, has been able to take advantage of this investment, working on an academic and industry task force to develop principles for the responsible use of generative AI in the law.

Harvey
Working with several law firms and consultancy group PwC, the OpenAI-backed legal start-up has grown to around 30 people since it was founded last year. Lawyers are not only using the tool for common tasks, including drafting and summarisation, but also in more creative ways to develop litigation strategies. They feed Harvey their arguments and then ask it to provide a counterargument. Where possible, the tool provides linked citations to improve confidence in the accuracy of what it produces.

Icertis 
The contract management software company is working with customers, including pharmaceutical company Johnson & Johnson, drug distributor AmerisourceBergen, and Boston Consulting Group, as part of an early adopter programme for its new generative AI tools, launched in July. The tools, which use a combination of Icertis’s AI models and OpenAI’s large language models, make it easier for clients to interrogate contracts and assess whether they meet company standards. For example, one customer was able to check payment terms in its customer contracts easily and put in place a project to reduce them — a change estimated to add $30mn to the company’s cash reserves.

Ironclad 
The San Francisco-based legal technology company made its generative AI tool publicly available to the market in April this year. To reduce the number of contract templates it was using, US-based gym company Orangetheory used the tool to consolidate its templates from 1,000 down to 250. The AI helped identify duplicate terms and drafted suggestions for revised language, which the legal team then reviewed. The team estimates the project was completed in half the time expected.

LexisNexis 
The data analytics business is introducing generative AI-powered drafting and summarisation — along with a search function similar to ChatGPT — into its legal product offering. The features are currently available to a limited number of trial customers, including law firms, in-house legal teams, and the US courts. To help address accuracy concerns, the company has included additional features, such as a prompt engineering feature, which suggests follow-up questions that a user could ask to elicit a more specific answer. LexisNexis has also launched a series of webinars and workshops on AI, which around 12,000 customers have signed up to so far.

Persuit 
Persuit, which acts as a marketplace for companies to hire external lawyers, introduced generative AI in the form of a “proposal analyser” to help in-house legal teams decide which law firms should be awarded work. Law firms submit proposals — which are often long, covering legal strategy, diversity commitments, and other factors for consideration — to Persuit’s platform. The AI creates a summary, comparing the proposals with each other. Around one-third of Persuit’s clients have opted to use the feature to date.

PwC Legal Business Solutions (LBS) 
Following its exclusive tie-up with Harvey in March, PwC announced in October that it was building a proprietary legal model with Open AI. The aim is to collect legislative data from around 90 countries using Harvey’s platform so that the firm can offer a legal knowledge subscription service to clients. Within PwC, Harvey is now being used by 6,000 professionals, of which nearly half are lawyers. In addition, the firm is also working with King's College London to offer a legal prompt engineering course, which teaches people the best kinds of questions to ask large language models. PwC also has an alliance with ContractPodAi on its generative AI product Leah, which it uses with clients.

Robin AI 
Founded in 2019, the company has built an AI “co-pilot” to help businesses manage contracts more quickly and easily than using traditional tools. It was the first to partner with the OpenAI breakaway Anthropic to develop its large language model Claude to be used by the legal sector. Robin AI combines older forms of predictive AI with generative AI in order to draft, review, search, and raise queries on contracts. It now has more than 60 customers, who are a mix of corporate legal teams and asset managers. A free version of the tool is also available for anyone to use on the company’s website.

SirionLabs
The legal technology company was one of the first to use AI technology in contract management. Using its own large language models, of which there are more than 300, the company allows clients to search a contract in a safe environment without the risk of confidential data being leaked to third-party clouds. A collaboration with IBM Watson on its new offering, called ‘WatsonX.Data’, means clients can put their data sets into a “studio” — a private, secure place in the cloud — and access different large language models. IBM is Sirion’s “client zero”: Sirion handles IBM’s sales and procurement contracts, and the two companies will market the safe studio together to corporate clients.

Spellbook
One of the earliest generative AI companies in the legal sector, Spellbook has been integrating the technology into its product since September last year. Focusing on contract drafting and review, the technology gives lawyers real-time recommendations as they draft in Microsoft Word. The technology leverages multiple large language models, including OpenAI’s GPT-4 and Anthropic’s Claude. Spellbook is used by around 1,000 law firms and in-house legal teams globally, and the company says it is currently taking on an average of 70 law firms per week.

Thomson Reuters 
The legal data and media company has taken a “build, partner and buy” approach to integrate generative AI as quickly as possible into its products, to offer services such as summarisation and an enhanced search. In June, the company paid $650mn for 10-year-old legal AI business Casetext, which specialises in generative AI and created one of the first “legal assistants” on the market. Earlier acquisitions include legal database Westlaw and template library Practical Law. The company is also working with Microsoft to develop a generative AI drafting solution powered by Thomson Reuters’ content and Microsoft’s AI assistant “Copilot”, which will be available in Microsoft Word.

273 Ventures
From consulting to designing products, the legal tech consultancy made headlines this year when it led a test of generative AI’s capabilities by arranging for GPT-4 to take the US bar exam, in which it scored in the 90th percentile. The company advises law firms and corporate legal teams and invests in legal technology companies. It developed the YCNBot with UK law firm Travers Smith, which allows users to safely run ChatGPT-4 on their own data without compromising confidentiality. The team also created Kelvin Legal Data OS, a data set and operating system that includes foundational legal, regulatory and financial text, which legal teams can use to train their own large language models.