We spoke to founder and chief executive Lewis Liu and general counsel Adam Eastell about Eigen Technologies’ ambitious plans to classify and extract data across the firm, across all practice areas, and across the front, mid and back office. If you didn’t catch this in the free monthly Orange Rag first time round, click here to register:http://legaltechnology.com//latest-newsletter/
Eigen Technologies in June announced that it has completed a £13m ($17.5m) Series A funding round co-led by Goldman Sachs Principal Strategic Investments and Temasek, and for many people in the legal sector it was one of the first times they’d heard of the London-headquartered company that counts Linklaters, Goldman, Evercore and ING among its clients.
Founded in 2014, Eigen’s natural language processing technology automates the extraction of unstructured, qualitative data. It has 50 staff based in London and New York and began gaining profile in the legal sector towards the beginning of 2017, when it won Linklaters as a client. The Linklaters IT team and lawyers collaborated with Eigen to launch what we described at the time as a homegrown AI-backed data analysis and extraction tool called Nakhoda.
Liu told us: “Linklaters was our first client but we rapidly moved into financial institutions and have a wide range of banks like Goldman and ING and banking clients in the US Asia and Europe.
“Only Goldman and ING allow us to name them publicly but imagine those types of banking clients and that’s what we have.”
Eigen works across the front, middle and back office, working with legal teams as well as with bankers to help them automate banking transactions.
Liu said: “We work across the entire spectrum of an organisation: the idea of the platform is that it is industry and language and document agnostic.”
The same applies in the legal sector where, while Linklaters is the only official law firm client, Liu says Eigen is working with other global law firms of Linklaters’ calibre. In November 2017 Eigen appointed Slaughter and May partner Adam Eastell as its first general counsel, and Eastell is bringing his experience to bear in how the product is developed and marketed.
Eastell tells us: “To achieve user engagement you do need to understand how law firms work and you do need to work with the clients to come up with a thoughtful approach to using the technology. You can’t just roll it out in a law firm and that’s that.”
However, he adds: “I’m very conscious at the end of the day that we’re extracting information from contracts – the workflow that goes with that are applicable to whoever is extracting that information, whether that be law firms and bank.”
The tech startup or scaleup now has plans to enter the insurance markets and works with top management consultants. For a small company to work in both the financial and legal sector alone must be a stretch leading one CIO to ask ‘How can a small company retain a focus across multiple industries? Does it not make most sense to specialise?’
However ,Eigen is bullish about the scalability and instant usability of its tech. Liu says: “We are working with a legal department in a bank and we have a larger scale legal AI mandate where we’ve analysed every single finance agreement that the bank has ever entered into and we have extracted information across all of those documents. That’s a scale I don’t think any of our competitors have done.”
Part of its optimism stems from the fact that the users themselves can train the machine. Eastell says: “The user is the person that asks the question, trains the machine and runs the extraction. Over time you have users who will be super users and can train others. For that reason it is highly scalable.”
Liu says: “The way to think of it is not to think of as a platform that does a task but what one of clients called an ‘instantly scalable workforce.’ Any non tech user, for example, a lawyer or human resources manager can take the machine and give it a couple of examples and teach it to answer a specific question, for example ‘what is the prevailing law’ or ‘is this release about buying and selling shares’: we require less training data so the user can very quickly train and identify the task it has to do.”
Eastell says: “The other point is that it doesn’t just pull out the clauses for humans to review, it gives you the answer, for example ‘what is the license fee’ – it gives you the answer. Or ‘what is the governing law’ – it doesn’t just give you the clause, it says ‘England and Wales’.
Liu says: “For a complicated document you might need 30/40 training documents but for something simple you might only need two to get started.”
He adds: “The really key point here is that we have a lot of scientists and mathematicians and will continue to heavily invest. We haven’t just taken a random machine learning model and built a sexy product around it. We have a way of modelling language that is extremely unique in certain mathematical forms, enabling our machine learning model to be very fast to learn.
“We have been very thoughtful about the mathematics we’ve deployed and extremely targeted about how you build a product that is user driven. A lot of our competitors need an engineer to come in and code. Ours is all about the user teaching the machine itself.”
Eigen also has a full suite of APIs and for a couple of banking clients, this means that any time an agreement is issued it gets pushed into Eigen’s system, where it is used for risk modelling and to assist and educate the inhouse legal team in terms of how to negotiate.
Liu freely admits that the technology is still immature and key to making progress going forward – and this is in part where the money from Goldman and Temasek comes in – will be investing in further top level technical expertise.
He says: “We have made it clear that we’re just starting up: we have great wins and we can do things with machine learning that no-one can do and we’ve made clear we will be doubling down on hiring in PHDs and NLPs and the right type of mathematicians.
“The dream that Adam and I talk about is when we can interrogate every piece of qualitative data: we’re just getting started. At the click of a button you could ask the machine ‘hey I want to know what the mood of the company is today.’ Or ‘what is the average notice period?’ That is our vision. We do have a way to go to put that together but we have the right seeds to build that technology and that’s why Goldman invested in us.”
It is true that for two such prestigious investors to inject this level of funding in a Series A is unusual and you can rest assured that they crawled all over the technology.
Eigen will now use some of that funding to invest in marketing, press and sales. Seventy percent of its staff are technical and Liu said: “Generally a lot of people do wonder why we have been under the radar so long. We just wanted to get the product and the technology right. We don’t want to disappoint our clients. We want to set realistic expectations. We’re just starting to invest in marketing and PR and sales. For the longest time didn’t have a sales team, we were just focussed on the product.”
The announcement of Goldman and Temasek’s investment has already put Eigen firmly on the map and Liu says: “Since the announcement we have already seen a massive spike in interest. Take the top 50 global law firms in terms of revenue: around 50% have reached out to us in the last two weeks.”
The use cases within law firms so far have been varied, from those who use Eigen to analyse their invoices, to their precedent banks and Eastell says: “Precedents are one of my big bug bears and I would love to break down a precedent bank and by doing this produce a precise M&A due diligence tool as well as intelligence on how people contract; what provisions are normally agreed; what percentage have break fees. I’d love to see that information broken down and analysed and made available to lawyers on a transaction, so if a lawyer doesn’t know something, instead of walking to another office for a chat they can get the definitive view. There is a huge amount of information waiting to be analysed.”
One US law firm is using Eigen to analyse how the terms of its fixed instruments have changed. “I can ask Eigen ‘in the last 100 transactions, is this a deviation from the market norm,’” says Liu. He adds: “We want to be the piece of technology that is used across the entire firm: by the back office; by the knowledge management team and by lawyers.”
These are big ambitions but with £13m in its pocket and the backing of Goldman, Eigen has the potential to be truly disruptive.
We also put a few questions from the market to Liu as follows:
Do you expect to create specific legal products and if so how will you get traction in a crowded and immature market?
There are a couple of angles to think about. Do we plan for our platform to be specifically for lawyers? Yes and no. There will be features for lawyers incorporated in our general platform including being able to review multiple documents at once. The platform has specific features that our legal clients have requested and Adam has provided guidance and those product features will be in our platform or already are. But those features could also be useful if you’re an investment banker looking after a collateralised loan portfolio. The point is that our legal clients drive a large part of our functionality but that doesn’t mean that functionality is exclusive to our legal clients.
How will you encourage firms to put sufficient lawyer time into products to make them viable / how much can any solutions be developed without specific legal input?
We do have a lot of input from legal clients. Keep in mind that within banking clients are the in-house legal teams. Obviously Linklaters is a client and we have Adam as well. We quite like being a second mover in the legal space and we can learn from the mistakes of those before us. Most of the successful Silicon Valley players are second movers. Facebook learned from MySpace. Google learned from Yahoo, which created portals that didn’t scale. Google came along with a general engine that scaled up and became extremely powerful. It does take a lot less time to train our machine. You only need five minutes to get something started. The ability of the machine to learn quickly is one of the biggest hurdles. So much time is taken teaching the machine of our competitors that people feel like they can’t be bothered.
Do you see a time when this is just plug and play and requires very little training?
There are two modes of operation for the platform. Our API platform is not plug and play. When a banking client needs to pipe their data through to Eigen and back you need to get engineers to control that and probably always will. Then there’s the product out of the box. You can push a button and get an instance in five minutes, we have that capability. But unlike our competitors we don’t have a set of pretrained questions yet. Luminance or Kira have up to 50 questions where you don’t need to teach the machine. When you get the Eigen platform it’s generally a blank slate. You do need to teach the machine. That being said, you can ask it any question you want and we’ve seen big law firms changing to our platform because of that flexibility. We do have plans to put in pre-trained questions and that’s in our roadmap.
How can a small company retain a focus across multiple industries e.g banking and legal. Does it not make most sense to specialise?
For us the focus is on solving the problem of natural language extraction. I look across our technical staff and 25-30% have PHDs and that is the problem we are solving. If you are literate and able to learn you can go from industry to industry. You do need specific industry knowledge to understand the pain points of an industry but that lies more in the sales side than the core technology. The technology is the most difficult piece to get right.