Many legal teams are still grappling with how to deploy the fast-growing AI technology. (iStock.com/Thapana Onphalai)
It took decades to persuade clients to accept law firms moving systems to the cloud, with some still describing themselves as "cloud‑averse" even though they use cloud services themselves. Artificial intelligence, meanwhile, has gone from "experimental" to "expected" in what Goodier called "almost a nanosecond," she told a forum run by Harvey, its first, on Feb. 23 and 24 in London.
"You really don't have time now to sit back and wait — you have to engage with this whether it's difficult or not," she said.
There's also the question of betting on the right horse — or horses — as firms weigh whether to build, buy or partner on legal AI.
Firms like Allen Overy Shearman Sterling were early adopters. A&O Shearman announced a "strategic" partnership to roll out software from the U.S.-based Harvey firmwide in February 2023.
But Sweden's Legora, one of its biggest legal AI rivals, has since made notable inroads, including firmwide deployment at Linklaters LLP and a global rollout at White & Case LLP announced in late 2025.
And, earlier this year, Anthropic's release of a legal plug‑in for its Claude product upended the market, prompting a global selloff in software and services stocks.
"It's a bet, and you need to be agile," David Wakeling, A&O Shearman partner and global head of the firm's AI advisory practice, said at the Harvey event. "When you see what Google's doing next, when you see what Anthropic is doing next, maybe there are competitors to Harvey — you never know where things are going to go."
Harvey is now a "multimodal" platform, he noted, meaning it no longer relies on a single underlying AI provider in the way it did at the start when it used only OpenAI's ChatGPT technology.
But when Wakeling first got involved, Harvey was essentially just two young founders working out of an Airbnb — they did have OpenAI's backing but nothing more in terms of infrastructure and track record.
"I wanted to check they were in it for the long haul," he said of his visit to OpenAI's San Francisco offices, where he and Harvey's founders met OpenAI's leadership. "We needed to make sure there was operational resilience."
That trip, after which he was satisfied that OpenAI had backed the founders and followed through, has shaped how he now thinks about new legal AI investments.
Wakeling said the decision on whether to build a new legal AI tool in‑house, buy an existing product or partner with a provider turns heavily on "portability" — avoiding being locked into a single model provider — as well as the cost and risk of maintaining in‑house systems that may quickly become obsolete.
"You're baking huge amounts of knowledge into a system, and then something better comes along, and you're kind of stuck, you're captive to an obsolete system," he said.
That's why his team insists on keeping their systems flexible enough that they can switch away from any one vendor if they need to.
"We still happily build Harvey workflows, but we have a plan of portability, which maybe keeps Harvey on their toes," he said.
He suggested that firms should only build the parts of an AI system that contain their own special know‑how, and they should build those parts so they can be moved or adapted easily as the underlying technology changes.
He also encouraged firms both to work with legal tech providers and develop some things in-house.
For example, the firm has built a system called ContractMatrix in partnership with Harvey and Microsoft, which is an AI-powered contract drafting and negotiation tool that is capable of connecting to whichever underlying AI model the firm chooses.
"The concept is we can unplug an AI system and plug in a new one, and we've still got everything hosted at the top," he said.
Emma Dowden, Burges Salmon's chief operating officer, said that historically the firm has preferred to buy and configure technology rather than build it.
"We're a law firm, not a tech company," she said at the event.
Developing systems in‑house can be a distraction that carries significant risk and cost — "millions of pounds" to maintain, she added.
But with generative AI, particularly agentic AI, "we are finding there is more need for build," she said. "You've got to have developer capabilities to customize tools around your own data stack."
Dowden added that in practice, this means a blend of approaches, with partnering remaining absolutely critical.
"It's fundamentally important that you pick the right partners, because all of our reputations are on the line now," she said. "It's picking the right tools and the right business partners."
UBS' in-house legal department began its AI journey by deliberately not fixating on the tools available in the market.
Barbara Koch‑Lehmann, chief operating officer of group legal at UBS, said the volume of AI tools on the market can feel "quite overwhelming." Instead of starting with technology selection, her team chose to step back and develop a vision and clear business outcomes first.
"We wanted to transform how we work, that's how we started, about a year and a half ago," she said.
UBS then asked staff to submit AI ideas from the ground up and compared those suggestions with its own strategic view of the business. That process generated around 16 to 18 possible applications, from which the bank narrowed the list to five priority projects — three aimed at its in‑house legal team and two focused on managing external law firms.
"For us, the hurdle was at least 140% ROI within a year, and that's then how we modeled our business case based on these top three and two use cases," she said.
UBS now has several AI tools in place, including Harvey and some built in‑house.
"We are looking at these AI tools to really augment our own people," she said.
This means freeing lawyers from routine tasks so they can spend more time on higher‑value work — such as providing strategic guidance when the business enters new markets or rolls out new products.
But Koch-Lehmann also noted that she was interested in using AI to save money on outside counsel.
And law firms are feeling the pressure.
Ashurst's Goodier said that even three months ago clients were still asking whether their advisers used AI, but now the question has shifted to how they use it. Clients have become more "sophisticated" in a short space of time and now want to understand exactly how the work is done, who is doing it, and which AI tools are being used, she said.
"We're having more and more clients that are asking us to quantify the savings from AI on the face of the invoice," she said. "But if you start to have a conversation with clients about the value that can be delivered through AI and how we do that, most of them are quite pragmatic and open to having that conversation."
When it comes to justifying spending corporate budgets on AI, Henry Gardener, chief risk officer at Markel Insurance, said it's hard to win extra tech budgets just for legal. So he pitched using Harvey in a highly "legal-adjacent" underwriting team that handled a huge number of merger-and-acquisition-related insurance deals in order to justify the cost.
"Anything that makes them faster, makes them more accurate, has a really big benefit," he said.
Adopting Harvey in February 2024 made the underwriting team so much faster and more accurate that they grew revenue from $19 million to $50 million in eight months, according to Gardener. Tasks that once took 40 hours a week now require a quarter of that, leaving more room to pursue other business.
"That's the leverage that we've been getting and seeing in pockets," he said. "It's now for us, how do we spread that out broadly?"
The blunt, "obvious" return on investment is cutting headcount, according to Damon Miño, a senior vice president at digital contracting platform Ironclad.
"That's the elephant in the room," he said.
But instead of selling AI as a way to replace lawyers, he encouraged GCs to focus on how it lets the same team turn contracts around faster, avoid extra hiring and process far more work than was possible before.
"That's a really easy story to tell," Miño said. "Think about the business outcomes that you produce for your colleagues, and how you can measure more return on whatever that is without reducing headcount."
But ultimately, both in-house teams and law firms will need to move quickly to adjust.
"This is coming at us like a freight train," Goodier said.
--Editing by Robert Rudinger.
For a reprint of this article, please contact reprints@law360.com.
