This article is a reprint of an Artificial Lawyer interview made in November 2016 and is part of a retrospective on legal AI and automation company founders published over the Christmas and New Year period. [Main photo: Ulf Zetterberg, CEO, Seal Software.]
Artificial Lawyer caught up with legal AI company founder and CEO, Ulf Zetterberg, during his recent visit to London. We discussed how his AI company, Seal Software, got going and just how massive an impact AI-driven contract discovery, analytics and review will have on the business world.
Artificial Lawyer meets Ulf Zetterberg in a co-working space on the South Bank in London, just behind the Oxo Tower on the River Thames. Zetterberg is in an energetic mood. There is a lot happening at Seal Software at the moment.
The San Francisco-based company recently announced Version 5.0 of its AI-driven document analytics system that allows users to operate it within Microsoft Word. The new V 5.0 also has improved customisation that allows clients to more easily train the AI’s machine learning functions. And yes, the company is already into the fifth iteration of its software.
While AI-powered document review may seem a new idea in some parts of the legal world, the reality is that some companies have been working in this field for quite some time. For example, Seal launched in 2010.
In fact, Zetterberg has been looking into digital data and information extraction for a long time. In 1994 he joined Proact, a European data storage business and by 1997 he was the company’s CEO. More software and data-focused jobs followed and by 2007 he was a Senior Vice President at OpenText, which recently bought e-discovery company Recommind and announced revenues of just under $2bn.
That’s an impressive background, but how did Seal come about?
‘The focus was not at all on legal at first. It was about document management. The problem companies faced was that they didn’t know where all their information was,’ he says.
Companies literally could not find vital legal contracts that they had signed. That was more than a bit embarrassing for executives, and was potentially business critical. Data search quickly became ‘a big thing’ and many other software experts saw the same need, spawning a myriad of enterprise search companies around the world throughout the 1990s. But that was not the end of the story.
Search sought to move beyond the mundane, and in many cases quite useless, keyword methodology. Limited semantic software was developed so clients could search for a whole sentence in order to retrieve the documents they wanted.
However, natural language processing (NLP) and machine learning then opened up a whole new opportunity. What if you could use ‘semantic’ software to actually read the entire document and tell you what it meant, not just find it for you?
Zetterberg noted that even when executives found the documents they were looking for, perhaps a contract with a supplier, very few wanted to read 50 pages of terms and conditions. They also didn’t always want to hand it over to the lawyers to tell them what it meant, whether inhouse or external lawyers.
That was a bit of a dilemma. If they didn’t understand a key contract of their business they were taking a major risk. But then, calling in the lawyers every time a salesperson or procurement head had a query about a contract would be both very expensive and very slow. Something had to be done.
Zetterberg and Kevin Gidney, Seal’s Co-Founder and CTO saw that if they could tap machine learning and NLP then they could solve this problem. Moreover they could develop a solution that would meet the needs of a huge swathe of the economy, not just help in the legal dimension.
With these ideas in mind Zetterberg and Gidney both left OpenText in the early summer of 2009. After a short interlude the company was founded and Zetterberg found himself once again the CEO.
For a company that is from one perspective at the heart of legal AI, Zetterberg is not solely focused on lawyers.
‘You don’t need to be a lawyer to create legal tech,’ he says with considerable credibility and in stark opposition to the views of some investors who have said you must have a lawyer as a co-founder to be a good legal tech start-up.
‘We don’t sell to law firms,’ Zetterberg continues, ‘the customers are typically not external counsel.’
As a company Seal has found that when they talk to big companies the feedback is that the executives there want to minimise their use of external counsel and want to rely on internal counsel to help develop the organisation’s legal and business strategy.
Zetterberg adds: ‘The big users of contracts are not only lawyers. Therefore, unlike our competitors, we are not built for only one sector of contract users. We provide insight into contracts for all within a corporation who may need access to contractual information and provided to them in the way they need to see it and as developed by them. Our users span all disciplines: procurement, compliance, legal, sales, human resources and others.’
That is to say, Seal serves the needs of the people who have to put all those thousands of contracts into action.
Clients include: Microsoft, Bosch, Dropbox, Experian, PayPal, Vodafone, Humana, DocuSign and even HP. The latter is perhaps slightly ironic given that HP bought UK tech company Autonomy, whose founder Mike Lynch has just helped to fund and create a new legal AI document review company, Luminance.
Seal also provides its software to Deloitte, probably the Big Four accounting firm that has most publicly acknowledged its interest in tapping legal AI systems. Deloitte also works with Kira’s AI software. Seal is also working with PWC and Accenture.
Today the company says the average number of documents a corporate client has ‘in Seal’ i.e. hosted on its servers for review across a number of areas, is around 100,000. There are five clients with more than two million documents hosted by Seal and one client may soon be moving to three million contracts in the system.
Perhaps one factor that differentiates Seal, aside from the stellar client base, is the way they partner with other global software companies so that their AI becomes integrated into the wider business world.
For example, they have a partnership with Salesforce the CRM company. Seal integrates with the Salesforcesoftware so that users can move seamlessly from the CRM functions of the system to using Seal’s document analytics capability within the same ecosystem, i.e. no need for a clunky switch to another application.
Usability is key for Seal, and the system was designed for business users to be able to extract contract intelligence for business decisions, and not just trained lawyers or data scientists.
Seal is also a little different in terms of the scale of funding. Zetterberg says the company has had over $30m in funding so far, including investment from Toba Capital, which is a fund set up by Vinny Smith, the founder of Quest Software, the IT infrastructure company.
Zetterberg has developed a long-term perspective on the software world and he is pragmatic about legal AI and its many start-ups. He notes that getting going as a start-up is not the biggest challenge. It’s proving to investors that you have sufficient clients and sufficient revenue to deliver ROI that is the challenge. For their part, Seal states that the company is seeing a 100% increase in client revenue every year.
‘Some of the new [legal AI] companies will become leaders, the rest will struggle to get further funding. It’s an issue of market share,’ he says.
For some start-ups that may be a controversial view, where many new ventures in legal AI see the sector as still very much ‘up for grabs’. But, it is probably true that at some point certain big legal AI brands will control a significant share of the market. Legal AI may be a new field, but it cannot defy economic laws.
On the subject of who will survive and who will not in this new legal AI landscape, Zetterberg points out that Legal Process Outsourcers (LPOs) are in a tough spot.
‘LPOs? This is low value work. If they don’t continue to strive to offset pure manual work through the use of technology, they’ve had it,’ Zetterberg concludes.
When it comes to lawyers and their dominant position in contract review, Zetterberg also does not hold back.
‘One of the biggest challenges to automation is the domination of certain professions, as they can be intimidating. You see this in healthcare, you see this with lawyers. But, once you get past the intimidation [of the clients by the lawyers] people should be able to understand what they need to do,’ he says.
Although not an advocate of ‘the end of lawyers’, Zetterberg certainly believes the automation of document review and other areas of inhouse legal activity could have a huge impact. ‘Its about empowering lawyers to be more effective and proactive when they support the business. Getting the right documents and right information into the lawyers’ hands more quickly is what give an organisation risk avoidance and competitive advantage.’
‘Out of the world’s $700bn total legal spend, the biggest part of that is corporate legal spending. If you automate [much of that corporate legal activity], then we could see about half the number of lawyers performing high volume, lower value work,’ he concludes.
Zetterberg adds that the problem all businesses face is handling ‘exceptions’. That is to say, an executive finds a contract and wants to deal with it, but how do they know what is unusual and exceptional, i.e. which one needs a real lawyer to look at it, and what is just the same old template filled with the same old legal clauses? If you can use AI to close the gap in exception handling then it will have a massive impact.
One might say that lawyers have a tendency toward complexity, while corporates are trying to move toward simplicity so they can get on with their business. Legal AI therefore will play an increasingly vital role in this relationship.
Another key issue is the use of data. Zetterberg notes that despite all the talk of Big Data companies don’t approach the issue with a comprehensive strategy. He recalls the example of a company that did a huge manual document review to find some information it needed. Some time later it had another query, but because they hadn’t anticipated the need for this new data, they had to repeat the entire process again. There is no way to ‘future proof’ a manual review, and so it cost a lot of money that didn’t need to be spent. took a lot of time and, quite clearly, it was incredibly inefficient.
One might say that such behaviour is like buying a car, driving it once and discarding it, then buying another one the next time you needed to get somewhere else.
‘When you combine knowledge management with AI, then you have something very powerful,’ Zetterberg concludes. Looking at the market as it stands today, then it is hard to disagree with Zetterberg on that point.
In fact, one would not bet against a worldwide revolution taking place in terms of AI-powered contract analytics and review in the coming years. And Seal seems very likely to be among the vanguard of that movement.