Lawyers’ skills will remain in demand, at least for the short term
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Nick West was inundated. Soon after advertising in March for a “prompt engineer” — someone skilled in presenting the right questions to a generative AI model such as ChatGPT — his inbox at law firm Mishcon de Reya received hundreds of applications.
But the chief strategy officer soon realised he was looking for the wrong thing. Despite various prognostications, the technology was not ready to be used meaningfully and safely by Mishcon’s 1,300 staff, let alone take the place of a lawyer.
“You have to ask yourself the question, in an industry that doesn’t really tolerate defects, how good is this stuff, how do we safely use it, how much could it replace a human and how much does it augment one of our experts?” says West.
He has been experimenting with setting large language models (LLMs) tasks such as summarising or drafting legal documents.
“I expected enterprise-grade tools to come quicker . . . I thought I’d be able to give my users something quicker, and therefore wanted to get hold of the kind of skill set that would allow them to use it better. It turned out it took longer to give them a safe way to use it.”
Now that Microsoft has announced its Azure Open AI service, which allows clients to run a GPT model in their own environment, easing concerns of data security, and Open AI has introduced its enterprise suite, Mishcon is experimenting in-house with different models. But, despite optimism among staff about the technology, West has found few processes that are “materially better” when delegated to generative AI.
Fairfax Insights, a consultancy, found in May that, after a few months of frenzy following the release of ChatGPT, parts of the industry were expressing “some scepticism and denial” about the impact of generative AI on law firms.
Yet talk of “workplace displacement” — a sanitised synonym for widespread job losses — was not without merit, Fairfax concluded in a report. “To the extent that more of the routine work, often handled by associates and paralegals, is replaced by AI, there will be an impact on the number of associates and paralegals required”. Even if lawyers are still needed to review the output, AI will put a premium on “skilled technology staff” who can get the best out of the tech, Fairfax added.
The spectre of a cull in legal jobs has, however, slowly receded.
Richard Susskind, technology adviser to the UK’s Lord Chief Justice and president of the Society for Computers and Law, says that although generative AI is a “remarkable development”, most of the “short-term claims being made about its impact on lawyers and the courts hugely overstate its likely impact”. The most exciting AI-fuelled developments, adds Susskind, whose doctorate in the 1980s was on AI and law, will not involve “swapping machines and lawyers” but will lie in innovations such as online dispute resolution.
Michael Gerstenzang, managing partner of New York-headquartered Cleary Gottlieb, warns that “lawyers . . . are predisposed to underappreciate the impact or potential impact of technology” and most senior practitioners “overvalue the things that they are good at”.
Still, he believes that “workforce displacement is among the least pressing of the topics” presented by generative AI and its adoption by law firms.
“Judgment still matters,” Gerstenzang says. Pattern recognition, the ability of an experienced lawyer to recognise the similarity of a particular case to historic litigation, “is capable of being very much delegated to [a] co-pilot”, such as a generative AI tool. However, “there is another aspect of judgment which is, ‘we’ve done this 100 times . . . but this one is different’.”
Such legal creativity, which can involve coming up with “a brand-new novel solution to a longstanding problem”, will not be replicated by a bot, says Gerstenzang. “We may get there, but today [AI is] going to naturally reinforce pattern recognition and prior ways of doing things.”
Joe Green, chief innovation officer at Gunderson Dettmer, which launched its own generative AI chat app in August, says that, while he expects it to “accelerate or replace” many tasks, “it won’t replace our people”.
“GenAI can give our subject-matter experts superpowers and move all our employees higher up the value chain, to focus on work that solves our clients’ business problems and is interesting and rewarding to do,” he suggests.
Mishcon’s West also sees generative AI as less of a job threat and more of a boon to business. “There are loads of things that don’t get done because they’re not cost-effective” such as reviewing all the service contracts of employees in an acquisition target, he says. Generative AI could do that work.
But the past few months have taught him not to be too confident in his predictions. “A year ago, I would have told you we were miles away from a tool that could draw a picture in the style of Andy Warhol,” he says. “So who knows what is coming?”