Lawyers face conflicting strategies for adopting AI

Are you a “fomo” or “foji” kind of person? These acronyms, denoting “fear of missing out” and “fear of joining in”, are more commonly used to describe awkwardness in social situations. But they could apply to how law firms respond to the latest challenges of artificial intelligence.

So, too, could “fud” — fear, uncertainty and doubt — as leading law firms grapple with how to engage with the threats and opportunities posed by last November’s launch of ChatGPT, the most prominent in a series of similar generative AI tools underpinned by computerised so-called large language models (LLMs).

Practice leaders facing confusion over whether, how and when to embrace the wave of new digital offerings must also consider the concept of the “hype-cycle”, as outlined by consultancy group Gartner.

This predicts a “slope of enlightenment” that typically follows an early stage of disillusionment as the benefits of a new technology start to “crystallise” and developers introd­uce second-generation products. Eventually, productivity increases as mainstream adoption unfolds.

Gerrit Beckhaus is a Freshfields Bruckhaus Deringer partner and now co-head of Freshfields Lab — the firm’s sizeable cross-functional team of legal professionals, software developers, machine learning engineers and project managers — but he cites his previous experience in the rollout of digital technologies as giving pause for thought.

In the past decade, before the emergence of large language models, other tools exploiting machine learning and artificial intelligence surfaced to spur law-firm digitisation. Around 2012, “the first wave of machine learning tools hit the market that basically had AI written all over them”, Beckhaus recalls.

Contract review and due diligence tasks were among that generation of tools’ first applications. Legal tech vendors claimed that “the machine would do your job”. However, their products and services “were not as advanced as advertised”, Beckhaus recalls. “The results were not sufficiently good,” even though “there were some ‘use cases’ that made life easier,” he says.

The hype, once recognised as such, triggered further scepticism. This was followed by the “trough of disillusionment”, as put by Gartner, that usually comes after overpromising by tech vendors pitching to law firms.

The scepticism that set in among law firms about the first generation of AI tools does not surprise Beckhaus. Lawyers “always aim for 100 per cent and therefore tend to overlook that, even if I get even 60 per cent done by a machine right, I’m further ahead than starting manually from the beginning”, he says.

By 2016, law firms began to accept they needed more tech experts to integrate digital tools into their work systems. By 2020, larger firms were attempting to integrate various applications on to single platforms.

That timing was crucial: Covid triggered more familiarity and comfort with digital tools among lawyers, like everyone else, which made them open to the latest wave of AI tools driven by large language models. “If [generative AI] had come five years earlier, the effect would have been different,” Beckhaus says. 

If generative AI had come five years earlier, the effect would have been different

Such commentary could give comfort to those in the “foji” rather than the “fomo” camp, willing to let others make the first moves in the arena. Yet, despite his historical observations, Beckhaus argues that OpenAI’s unleashing of ChatGPT on to the world last year will prove a watershed moment for the legal sector. “There’s the time before November 2022, and there’s a time after,” he says.

Headlines about ChatGPT and its availability pushed the digital transformation agenda at several law firms, Beckhaus adds. “Now, everyone can experience — and that wasn’t true before — what artificial intelligence means and what it could mean for the future.”

“Transformative” is how Dan Hunter, executive dean of the Dickson Poon School of Law at King’s College, London, also describes the launch of ChatGPT. “This is the most remark­able technology that I’ve seen in the AI space in 30-plus years working in AI and law,” says Hunter, who specialises in teaching tech and law and has launched his own start-ups focused on generative AI.

Tara Waters, chief digital officer at UK-based Ashurst, agrees a great change is under way. “We’ve been using AI tools for a number of years,” she says. But the latest large language models have created a “leapfrogging opportunity”, she adds.

The change has been abrupt. Funding for AI systems hardly figured when she was calculating her unit’s budget in February 2023. But, now, it tops her priorities.

Shilpa Bhandarkar, co-head of client solutions and innovation at UK-based Linklaters, says: “Until November 2022, it felt like all legal tech start-ups were selling AI — now all of them are selling generative AI.” But she expects any transform­ation to be steady, not fast.

AI is most beneficial to law firms when their documents and knowhow, which are deep banks of linguistic data, are “structured into a format that the algorithm can easily work with”, she says. Linklaters has its data “all on one secure platform and searchable for our people”, she adds.

The task facing law firms is to teach generative AI platforms with data they have controlled and checked for provenance, to reduce the odds of erroneous output — so-called “hallucinations” — produced by many generally available AI chatbots.

And, before lawyers delve too deeply into elaborate modifications for large language models, suggests Haig Tyler, chief technology officer at UK-based Herbert Smith Freehills, they should reckon with the sheer novelty of what is already available.

He compares large language models now to the invention of the automobile, and the success of Henry Ford’s Model T.

“You’ve got a really neat, basic model that we can do some cool stuff with,” he says. “[So] let’s not go to the custom shop just yet.”