This article is a reprint of an Artificial Lawyer interview made in August 2016 and is part of a retrospective on legal AI company founders published over the Christmas and New Year period.
Artificial Lawyer caught up with Noah Waisberg, Co-Founder & CEO of Kira Systems. We discussed how Kira Systems got started, how it is different to other AI systems and where it’s going.
Co-founder of Kira Systems, Noah Waisberg, used to be an associate at Weil Gotshal & Manges, starting there as a junior associate in 2006. A lot of his work focused on M&A and he says he found himself spending a lot of time reviewing contracts during due diligence (DD).
According to Waisberg, DD often amounts to between 30% and 60% of the legal bill law firms gave to their clients. In short, a lot of lawyers spend an enormous amount of time just reading through documents in search of information, often buried within clauses within the text.
Waisberg also noted how, at law firms right across the spectrum, junior lawyers often made mistakes during the DD exercise. And perhaps this was to be expected given that young lawyers and paralegals can work all weekend and long into the night, for example, in order to speed up a deal.
Despite the inefficiency, many large law firms had no problem with this approach as they had a huge number of junior lawyers and paralegals at hand that could be thrown at such tasks.
And although Waisberg doesn’t say it, it is well known in the legal industry that this inefficiency is very remunerative. Though, to be fair, if the client had a deadline there were, at least until now, no other ways of handling DD. Meanwhile, avoiding a thorough DD process was not really an option in the vast majority of transactions.
But was this ‘as good as it gets’? Couldn’t the due diligence process be improved?
Waisberg then started to notice that his fellow associates were always looking for the same kinds of information. Certain deals always drove a search for consent clauses, others to check the details of covenants, and property-related deals always created a demand to search for rent levels in the lease documents.
In short, there was a set of very clear patterns being repeated throughout the DD work. The information itself may have been locked away in potentially complex clauses, written in natural language, but surely there was a way of breaking that down and systematising this laborious process?
Scroll forward to January 2011 and Waisberg had left Weil Gotshal and teamed up with Dr Alexander Hudek from Canada’s University of Waterloo. Hudek had done work in the field of bioinformatics, which focused on finding similarities between DNA sequences. He is now the company’s CTO.
The Toronto-based team started by trying figure out a way to develop an applied AI system that could learn these search patterns and be turned into a program that lawyers could apply to their document searches.
Initially, the Kira Systems team tried to use open source, off-the-shelf machine learning algorithms. They pursued this strategy for around six months, but found the system never got accurate enough. The proto-Kira Systems software was limited in scope and could only consistently spot simple clauses.
A lot more machine learning R&D work was done, refining and improving the system and after two and a half years Waisberg says they were able to provide a compelling software solution that could deal with a wide range of information and types of clause.
The next job was to go and build up a client base, which the firm has done to some extent in a relatively short period. It publicly lists several law firms among its clients, including DLA Piper, Torys, Clifford Chance and Big Four accountant Deloitte. There are several others, including some relatively small law firms, though Kira Systems cannot for now mention all their names.
When asked about where Kira Systems fits into the emerging legal AI ecosystem Waisberg has a very clear idea of where his company stands. He points out that fundamentally ‘we are not trying to sell professional services. We’re selling machine learning contract analysis software.’
Kira Systems doesn’t stray into using its own manpower to help read documents or verify the results, nor does it seek to offer legal guidance on the information it extracts. Its job is clear: extract vital information from documents and do it with the accuracy and speed one would expect from an applied AI system, which can be a time saving of 20–90%.
He stresses that Kira Systems is somewhat different to some other AI companies in that the company simply licences their software to a law firm, or accountancy firm, and other than training to help them use the software, generally leaves the client to do what they want.
He notes Deloitte as an example. The accountant took the software and taught it themselves to search for over a 1,000 bespoke information points. Kira Systems was not involved in that search expansion, even if the Kira software was the ‘backbone’ for it.
Importantly, Waisberg notes: ‘I don’t know all the ways Deloitte uses our software. Our clients don’t necessarily tell us what they have done with Kira.’ While some may see this as a challenge, it could also be viewed as an advantage as not all clients want the know-how they develop when using AI to be passed onto other clients of the AI provider, no matter how unintentionally or indirectly. The ‘air gap’ between Kira Systems and what their clients do with the software ensures this.
In terms of how far Kira Systems’ applied AI can be used in new ways in a law firm, Waisberg is pragmatic. Rather than dream of far future applications, he instead focuses on immediate uses, for example using the software more widely across the many practices of a law firm, rather than just the one that has initially adopted it.
But he does see the software developing a greater ability to show the client where they may want to spend more attention in their DD work. That said, fundamentally, Waisberg does not seem to want to radically alter what Kira does, but rather to allow it to work seamlessly with other applications and software. In short: Kira works, so why change it? The job now is to see it used more broadly.
One area of growth Waisberg does believe Kira could develop further into is helping outsourcing/LPO companies improve both their efficiency and also the quality of their work.
He notes that even LPOs, which are usually considered to be cheaper than using a traditional law firm, also get ‘push back’ on the fees they charge. AI could help with that and improve accuracy levels for the end clients.
In terms of the business plan, Waisberg states that the company has no outside VC funding and end goals such as an IPO or other form of exit are ‘not something we are thinking about’.
With regard to how applied AI will impact the market as a whole, Waisberg notes that rather than become an area of legal tech that is the sole preserve of ‘Big Law’ it could in fact become ‘a great leveller’ that allowed smaller firms to do more and punch above their weight.
‘Quite often a small law firm cannot take on a piece of work because they don’t have enough people to do it quickly,’ he explains.
The key problem is not the advisory expertise of the senior partners, but rather that smaller firms cannot throw sufficient numbers of people to do the DD on a deal in a short space of time. That cuts them out from a lot of transactional work that otherwise they could perhaps take on.
Applied AI would help solve that challenge as it could greatly reinforce the DD capability of even a very small firm.
Artificial Lawyer then asked whether another barrier was cost?
Kira Systems has two main components it charges its law firm clients, one is the number of documents the client wants to read, and the other is how and where the software is hosted.
Waisberg was understandably not keen to give exact pricing details at present, in part because each firm’s needs would be different. However, from what could be gathered it is clear that even a small law firm in the UK 200 would be able to afford Kira Systems’ AI software.
While for the larger firms, with revenues in the £100millions and above, the cost of applied AI is relatively insignificant compared to the total fees such software would help generate.
Waisberg is candid about the process of bringing new clients on board. He notes that when a law firm considers using applied AI it naturally wants to evaluate the benefits. Larger firms have more capacity and internal tech expertise to do this, as compared to small firms. Hence, larger firms tend to be the ones making use of applied AI, for now, though this could change.
He adds that at the end of the day law firms are pragmatic, they are not using AI because it may be exciting or new. They want AI to provide a clear ROI, and, says Waisberg, it certainly does that.
‘Law firms are thinking: ‘Can we have a better business? What is the ROI here?’ And, as we have found, there is an actual ROI for law firms,’ Waisberg concludes.
It would seem then that applied AI has a bright future within the legal sector and Kira Systems seems destined to play a key part in this technological evolution.