Meet Sibyl AI – The New Claims Prediction System

Meet Sibyl AI – The New Claims Prediction System

This is a Guest Post by Richard StraussCEO and co-founder, Sibyl

Are we going to win? It’s a question clients love to ask, and lawyers hate to answer. It often results in a maze of fine balances, reasonable prospects and strong or weak arguments. But AI makes it possible to answer this question precisely, with a quantified margin of error.

There are really four things clients want to know when a new claim comes in:

(1) whether that claim is going to succeed;

(2) how much it will cost to deal with the claim;

(3) what the other side will settle for;

and (4) what the ultimate payout would be.

When you know those things, you can make quick and cost-effective decisions about which claims to contest, which claims to settle, and how much to pay.

At the moment, to get to that position, someone has to read through the papers and come to a view. We think that represents around 15% of the total claim handling cost on average (depending on the type of claim). And often, it’s not easy for a person to give clear answers.

Sibyl AI automates this part of the process. We predict claim outcomes instantly using machine learning. Our focus is on claims expressed mainly in writing – so legal claims are squarely within our market. But we’re also thinking about insurance claims and customer complaints in regulated industries.

So far, we’ve achieved 81% accuracy predicting whether claims relating to mortgage products will succeed. This dataset is essentially a block of text associated with an outcome, where a set of legal principles determines what happens.

It’s a good indicator of what our technology can achieve. But of course, we’re looking to demonstrate our capability for other claim types, with more representative data, and expect the accuracy of our predictions to increase over time.

The interest we’ve had from law firms has been really encouraging. Clients are increasingly expecting firms to deliver cost savings by making better use of technology. And firms are starting to see this as a way to expand their range of services, integrating more closely with clients’ processes.

Our Team

Sibyl AI has three co-founders: myself, a lawyer with a background in the City of London and the UK Government; Nikul Vagdama, an engineer at Oxford University with experience in the investment research industry (COO); and Sivo Daskalov, a doctoral candidate in machine learning (CTO).

Nik and Sivo met at Cambridge University, but the team first came together at last summer’s online courts hackathon, organised by the UK Courts and others. We won the prize for ‘coolest tech’ by predicting the outcome of employment claims, beating top law firms and tech companies. And that got us thinking, maybe we’ve got something here.

We’re a UK company concentrating on the UK legal and insurance markets. However, Sivo is based in Bulgaria, and that’s where we’re looking to expand our tech team (our first recruit will be joining in a few weeks). We’re fortunate to have access to the significant pool of tech talent in Bulgaria, and expect this to be a competitive advantage as we grow.

It’s been a steep learning curve for all of us, but we’ve got to the stage where we understand each other’s worlds, developing a process for selecting and testing powerful features based on legal insight, supported by software that allows us to iterate in minutes. We think this is the key to making superior predictions, providing clarity on questions lawyers find difficult to answer.

If you’d like to know more, here is a short video that gives an intro.

3 Trackbacks / Pingbacks

  1. Case Prediction Booming as Stewarts Pilots New Solomonic System | Artificial Lawyer
  2. Casecrunch Team Splits Up, CourtQuant is Born | Artificial Lawyer
  3. Big Data Startup Concirrus Wins Norton Rose InsurTech Prize – Artificial Lawyer

Comments are closed.