Start-ups are big business for Stephenson Law. Founded in 2017, the practice has ridden the wave of investment in emerging tech companies in Bristol, advising them on everything from privacy and procurement to employment contracts and intellectual property law.
Most of this work is exciting, explains founder Alice Stephenson. But as Bristol’s start-up scene goes from strength to strength, a troubling new trend has emerged among businesses who, unable to afford legal advice early in their existence, have let basic problems around contracts or privacy policies problems fester into their adolescence.
“Then they come to us, 12 months down the line, and we’re having to fix the problems that have been created by the bad advice,” says Stephenson. These problems can include unwitting trademark infringements or simple contract disputes. “It’s not the type of work that we want to do as a law firm,” says Stephenson. “We would rather work with these businesses and help them to be proactive, and get their businesses into a really good place, than helping them to fix mistakes.”
In response, Stephenson and her colleagues have used a £50,000 prize they won as part of the ‘Women in Innovation’ awards to develop a new interactive platform that provides start-ups with entry-level resources guiding them on a host of legal issues. Services include templates for employment contracts and privacy policies along with a chatbot capable of answering low-level queries about how the laws and guidelines contained in the database are relevant to their particular business.
The platform is an example of automation being used to tackle unmet demand for legal advice among consumers and SMEs. In November 2020, the UK’s Legal Services Board reported that half of the 31,500 small businesses it surveyed handled legal problems on their own. More than three million consumers per year encounter a need for legal advice that went unmet – unsurprising, given that 87% of those surveyed believe they cannot afford it.
Stephenson believes that technology will go some way in meeting this demand. So far, however, she has found little inspiration within her own profession. “There’s nothing, at the moment, within the legal industry that I’m really looking at and going, ‘That’s amazing’,” she says. That’s partly due to the sector’s traditional aversion to risk, she says.
Under the surface, though, the legal profession is slowly starting to shake itself out of the torpor of manual, paper-driven processes and embrace the automation of key background processes. This is an evolutionary process that should not only address that vast unmet need for legal services, but also create new roles that emerge to cement the role of technology in the profession, and help to change the role of the lawyer itself.
Automation and the legal back office
Before he founded his consultancy, D2 Legal Technology, Akber Datoo worked in IT at an investment bank. The lack of technological skill in the institution’s legal department was a constant source of frustration. “I [could] see the world moving towards systems and data, and I see my in-house legal team picking up parchment and a quill if they had the chance.”
The two decades since, however, have witnessed a sea change in the way the legal profession approaches technology. Advances in natural-language processing have boosted software’s ability to sift through and decode thousands of documents. This has led to the automation of a range of processes that would otherwise take human lawyers hours, if not days, creating a library of off-the-shelf programs that have become a familiar feature at many practices. These include tools aiding in legal discovery – the process by which documents relevant to a case are sourced and sequestered – to contract automation, legal research and due diligence.
Lawyers, who are the most conservative, with a small ‘c’, profession, are suddenly all ‘innovating’.
Joanna Goodman, author
The space is attracting increasing investment: between 2016 and 2019, investment in British ‘lawtech’ start-ups leapt from £2.5m to £62m. In that time, adoption has advanced significantly, says Joanna Goodman, a journalist and the author of Robots in Law, her treatise on artificial intelligence (AI) in the legal profession. It was published in 2016, “when there were no books about legal AI, and everybody told me I was writing a lot of hype and rubbish”. Now, though, she says: “Lawyers, who are the most conservative, with a small ‘c’, profession, are suddenly all ‘innovating’.”
That is increasingly borne out in the data. According to a survey of 353 lawyers published last year, some 80% said that they were familiar with document or knowledge management software. While 48% said that they used applications incorporating AI in their work. “We [found] that around a quarter of respondents are using it in the most popular context, which is for legal research,” says one of the survey’s authors, Professor John Armour of Oxford University. The next most popular applications were due diligence (16%) and e-discovery (12%.)
One barrier to adoption may be the need that AI presents for closer collaboration between technology and legal staff than for typical application deployments. “The system itself doesn’t just sort of drop-in out of nowhere, ready to run,” explains Armour. “It needs to be designed, and set up, and quality controlled, and trained, and managed.” But working practices are evolving to support this collaboration: 40% of respondents to Armour’s survey work in multi-disciplinary teams of lawyers and IT specialists.
Ninety per cent accurate information about 2,000 contracts might actually be more valuable to the client than highly accurate information about 200 contracts.
Oliver Campbell, Hogan Lovells
Multi-disciplinary teams have helped international law firm Hogan Lovells not only adopt AI solutions but also develop some of its own, including a solution that helps clients automatically update contracts in preparation for the retirement of the LIBOR benchmark interest rate. Innovating with ‘lawtech’ requires a range of capabilities that law firms may not traditionally have developed, explains Oliver Campbell, the firm’s head of practice operations.
Clients have responded positively to Hogan Lovells’ embrace of AI solutions, says Campbell, although the firm’s own lawyers have sometimes needed convincing. “Lawyers are trained to see risk,” says Campbell. Hypothetically, they may baulk at a contract analysis tool with only 90% accuracy. But what these tools may lack in accuracy they make up for in scalability, he adds. “Ninety per cent accurate information about 2,000 contracts might actually be more valuable to the client than highly accurate information about 200 contracts,” explains Campbell.
The number of multi-disciplinary teams will continue to grow as the adoption of legal automation services increases, Armour predicts. This is likely to make lawyers more productive, as mundane aspects of casework become mechanised. New roles may also emerge to help bridge the gap between technologists and legal professionals. “I think at this point it’s an open question whether we will call these people ‘lawyers,’ or whether they might have some other kind of appellation,” says Armour.
The limits of legal automation
Greater adoption of legal AI could, argues Armour, help to bring down the price of legal services and help address the vast unmet need for advice. Currently, however, not all firms have equal access to the technology.
“I’d say the future is here – it’s just unevenly distributed,” says Goodman. “The ‘Magic Circle’ firms can afford to buy all the products, and they can afford to hone them to do what they want.”
The same goes for commoditised practices, such as those working exclusively in insurance, firms whose data is more extensive and uniform. Mid-sized practices, such as regional partnerships, may take longer to embrace change, particularly when adopting AI. “You need to have the time and resources to train them, and not all mid-market firms have that,” explains Goodman.
Effective legal automation requires investment and training. Traditional firms who are making healthy profits without it may not feel the compulsion to pursue either. Additionally, “the career progression opportunities for people working in those teams who come from a non-legal background may be seen as constrained”, says Armour.
Legal automation is also not without its risks. Even the best-trained algorithms make mistakes. Goodman and Datoo both cite a case in which an AI-powered due diligence tool misinterpreted the word ‘ALL’, commonly capitalised in contracts, as a currency. “‘ALL’ just happens to be the ISO code for Albanian Lek,” explains Datoo. Rules for resolving disputes that arise from automated legal decisions, meanwhile, remain a work in progress.
But perhaps the most significant constraint on the pace of automation is the availability of training data with which to develop reliable AI models. Unlike other sectors such as financial services, where voluminous training data exists for AI models in the form of numerical datasets, the bread and butter of the law is the written document, often containing text suffused with impenetrable legalese and containing information on clients or companies with various privacy requirements.
There is also a ceiling on what NLP systems can understand about legal documents and convey back to the reader. Even language models, such as GPT-3, are not consistently reliable. Bots that can provide individualised legal advice therefore exist at the very frontier of legal AI, says Armour. “There’s just a very small level of adoption so far.”
Reducing the burden of admin
Even without this level of sophistication, automation can reduce the considerable administrational burden of providing and receiving legal services. Before joining Hogan Lovells, Campbell volunteered at a legal advice clinic, where most inquiries involved “helping people navigate forms and processes that seemed alien to them”.
This can be achieved through questionnaire-based systems, that are already helping to streamline estate planning and chatbots that dispense advice at a lower level of legal risk. An example of the latter is being jointly developed by AY&J Solicitors and the University of Bradford. The ‘Legal Immigration Artificial Intelligence Advice,’ or LILA, trains itself on documents on statutes, case law, and experts on immigration law to help navigate new arrivals through the UK settlement process. The ultimate goal, its developers say, is to accomplish up to 30% of an immigration lawyer’s work.
Similar systems could be applied to other areas like family law, helping to brief solicitors on the context of divorce proceedings without recourse to lengthy in-person interviews, says Goodman. These relatively simple solutions could also provide practices outside the ‘Magic Circle’ with a taste of automation and its benefits, she adds.
As it has for all industries, the pandemic has forced the legal profession to adopt digital tools faster than it might otherwise have done. And, as the chief technology officer of BCLP Cubed, recently told the Financial Times, recent progress with automation has reassured lawyers that “it’s a helping tool”, not a threat to their jobs.
Nevertheless, uptake is likely to continue to be gradual, not least because digital innovation is not the primary quality that clients seek in their lawyers. “I don’t think that everybody is looking for a lawyer that’s innovative,” says Goodman. “They’re looking for a lawyer that is effective.”