As one of the first 100 employees at Google, Ron Dolin worked directly on the indexing and retrieval of information on a massive scale. Before that, he was a physicist working on data acquisition and computer controls at European particle accelerator CERN, where scientists explore the fundamental laws of nature that create and destroy universes. Now the physics and computer science have come together for Dolin in a surprising place: the legal profession. Dolin doesn't think the traditional structure of the legal universe can survive what's coming from the world of artificial intelligence.
"These firms are looking at future doom if they don't start playing with business models in different ways," said Dolin, Instructor of Law at Harvard Law School.
Specifically, the cash-cow model of elite law firms — first-year associates racking up billable hours from endless hours of M&A contract document review, with the revenue flowing up the pyramid to partners — is facing an unprecedented challenge.
Artificial intelligence software can do the contract review work of first-year law associates at a speed and scale that no human could — or should — be asked to do. And as A.I. systems become more common, it becomes more problematic for law firms to not follow that trend.
"This is not reversible," Dolin said. "The first-year associate as cash cow to partnership is breaking."
It's not your typical rise-of-the-machines problem. In manufacturing scenarios, the robots lead to lower costs and better returns — often at the expense of jobs. In the legal world, becoming more efficient doesn't necessarily mean less lawyers — though no one knows for sure — but it does mean less billable hours and potentially less revenue for the partners.
"GE is selling a product for a set dollar amount, so the more efficient they can be and the more they can lower cost, the more margin they get. That is not the case with law firms," said Noah Waisberg, CEO and co-founder of A.I. software firm Kira Systems, who was an M&A lawyer before starting Kira. "With professional services firms, it's tricky to sell A.I. to them because unless they work on a fixed-fee model already, [Wachtell Lipton Rosen & Krantz is the prime example of an elite firm that doesn't use billable hours], if they are selling hourly, they really need lots of hours of output."
Law firms don't have their heads in the sand, but they would prefer that the partnership model makes some tweaks rather than being upended.
"Lots of clients are questioning billable hours," said Isabel Parker, director of legal services innovation at U.K.-based law firm Freshfields, which is one of the oldest law firms in the world (it has been the legal advisor to the Bank of England since 1743). "It's not just law firms but firms like Deloitte and EY. There is need for more price predictability and fixed fees, and automation allows you to run much more accurate analytics on the work you are doing," Parker said.
She said the pressure from corporate clients is an extension of what they are experiencing in their own worlds. "Big financial institutions face being put out of business on a daily basis, and they expect us to be looking at our own sector," Parker said. "It's not the death of the law firm partnership model, but it will change over time. No one can say how, exactly, but we have to think about how we price and work."
Matt Kesner, recently retired from his role as chief information officer at Fenwick & West, one of the top tech-sector law firms, said 30 years ago in his first job as a trial lawyer, he was working with "mini computers" — precursors to PCs — in the attempt to track vast amounts of data for trial discovery phase purposes. In the United States, at least, that role for lawyers was long ago outsourced and automated by e-discovery software — either because law firms were far too slow to adapt, or the work wasn't important enough to defend as part of their business. But the past five years has seen the encroachment of A.I. as a more fundamental challenge to the overarching model of many law firms.
Law and professional service firms are turning to start-ups like Kira Systems — and many others operating in the A.I. software niche — that can handle the contract review work of associates faster and more accurately. Deloitte's implementation of the Kira Systems technology last year, which is now being used by more than 5,000 employees, is the largest A.I. deployment at any professional services firm in the world, according to Kira Systems' Waisberg, which counts Deloitte, Fenwick, Freshfields, Latham & Watkins and Davis Polk among its growing roster of legal and professional services clients. Its system is now in use at four of the top 10 legal firms in the U.S., and five of the top 20 accounting firms. Internationally, its AI now reaches clients in 40 countries including Canada, the UK, Australia, India, and South Africa.
"Most corporate clients want the same efficiency from us that they want out of manufacturing partners," Kesner said. "This [A.I.] is one way to get there. It doesn't necessarily help us make more money. We may make less, but that's OK. It lets us focus on things more important to clients." He added, "I know it will sound pollyanish, but we really are trying to please clients. For us working with tech companies, it's the expectation. We're amazed and delighted with what they come up with, and for us, to not use it is silly."
Dolin noted that Google has a dashboard that allows its legal counsel to monitor the billing codes being generated by outside law firms.
"The leading indicator of change by far is data gathering and data acquisition. All the data is there and is increasingly enabling competition and pushback in a strong way. If you are charging me three times as much as someone else, can you tell me what is unique about this deposition? In-house corporate counsel has never had that power before but do now," Dolin said.
A.I. technology can review contract documents with the same or greater accuracy than lawyers in anywhere from 20 percent to 90 percent less time, according to Waisberg.
Freshfields' Parker said since it began using Kira in the fall of 2016, it has seen efficiency gains of up to 70 percent. Though AI has its own knowledge gaps: If the software encounters a document set or provision it has never seen before, the efficiency gains can be significantly lower, Parker said.
But legal AI is advancing rapidly. In just the past year, the market for sophisticated law firms has moved from buying AI solutions for particular problems (Kira for due diligence, for example) to creating end-to-end solutions by integrating different kinds of legal technology – extraction, expert logic and document automation, for example.
Parker said Freshfields is developing an AI solution for a large mass claims case which involves processing over 6,000 international claims with up to 200 additional claims each week. Parker explained that the vision is to use Kira to extract the information about the claim and claimant; auto-populate a claims management database; then use document automation to draft the submission to the court (automatically pulling information from the database about the claim that has been extracted by Kira, then automatically populating a template pleading); and then use expert logic to determine whether that case can be settled (applying a decision tree type structure — simple legal reasoning with a number of different variables).
"When you have a very large document set, this kind of simple automation can drive real efficiency for the client," she said.
First-year law associates aren't panicking. "Its ridiculous to think associates should do this over and over to just generate money," Dolin said. "They hated doing that work in the first place."
The law firm executives agreed that first-year associates are more relieved than worried. "These are smart people. They get frustrated doing that work over and over again, and clients don't want to pay for it. We can't sustain a model where humans are doing this work."
"The history of this has been different for every industry and in some tech eclipses what people do, and in other industries it is just an assist or supplement," Kesner said. "I don't think it is clear what will happen in law, and I don't think law is just one silo of jobs." He said trust and estate lawyers are already using software for a much bigger part of their jobs than a decade ago. "The expectation is they will have expensive software, and most work is done by computer. It isn't that way for corporate lawyers yet."
But there are no guarantees. Kira's main use at Freshfields is in its Manchester-based legal services center, where tasks defined as "repetitive legal work" are handled by paralegals. "Their reaction so far has been positive, because the parts of work taken by Kira need a lot of training. You can't just stick it in a room and expect the result to come out," Parker said. "That's a new set of roles for people that they didn't have before. ... A hybrid role between lawyer and tech."
But Parker added, "It's very difficult to predict, and it would be naîve to say it won't change the shape of our traditional workforce."
Waisberg said an important goal of Kira is to ultimately convince law firms that A.I. will lead to additional revenue opportunities. A typical M&A deal of $500 million might include review of 75 to 300 contracts. But those firms in total may have thousands of contracts between them. "Right now buyers only look at a small part. "They don't feel like paying fancy law firms thousands per contract to review all of them," Waisberg said. But A.I. holds out the possibility of increasing the scope of contract review. And increasing scope doesn't necessarily reduce the number of people, or the bill.
The law firms are reserving judgment on that. "There is no way of knowing whether Kira is right about the capacity to review more data," Parker said, because they do not know what would compel clients to ask for more data. But she does believe that the interaction between A.I. and lawyers creates opportunities for new types of premium pricing. Customizing the software — which the Kira system allows — to searching for the right terms has to be done by a qualified lawyer.
"That is a skill set that commands a premium. As a package, I can see how it will be more cost-effective to clients, but it's serious legal work and you can charge for it," Parker said.
Dolin — who now teaches courses in legal technology and informatics at Harvard and Notre Dame — said billable hours are only one problem within the larger problem of how legal partnerships think. They don't think like corporations, where a portion of revenue is paid out to shareholders in the form of dividends and the rest of the money put back into R&D and systems. Partners are both the corporate managers and shareholders.
"For some reason, partners think it's taking money out of their pocket to invest in new systems," Dolin said. In addition to his academic work, Dolin is an angel investor focused on legal technology start-ups.
Dolin does not worry about associates losing their jobs. It's not the model of a next generation of good attorneys that is breaking, he said. "But if you have a pyramid model, that's in trouble." He said firms are narrowing that model into a pipeline and continuing to make a profit that way, switching from pure billable hours as the dominant model to some fee structure that incentivizes efficiency and value.
A start-up like LegalZoom, which can now handle highly commoditized legal work, is not an immediate threat to elite law firms. But what happened within the U.S. market with e-discovery is cautionary. "That's an area where work has been disaggregated from the firm. That market has been taken, and it's hard for any law firm to compete there," Parker said. "We have to be alert that disaggregation can happen elsewhere."
Waisberg says that as AI like Kira Systems has become more mainstream and received more use at offices among the legal industry's big players, concerns about losing billable hours have decreased. "Firms have begun embracing the technology as an opportunity to do more and different things for their clients that just weren't possible before."
Only one thing is certain. "Firms resting on their laurels will be completely gone," Parker said.