Bitesize InsurTech: Tyche
March 11, 2017 George Hanks
Over the next four weeks we will be covering ‘Data and AI’ startups. This week, Tyche.
Tyche (Tie-kee) is a US based platform that uses data augmentation and artificial intelligence to help insurers and reinsurers better assess commercial casualty risk.
How it works
Insurers provide details of the risk e.g. company, location and industry. The Tyche platform interrogates various data sources e.g. Google search results. It takes this unstructured input, uses natural language processing to categorise words within the text, compares these to its internal model of relevance to claim rates and, finally, creates a claims likelihood model.
For example, an insurer wants to insure a hospital for workers comp. Typically insurers would know basic information such as ‘200 physicians work in this hospital’. Tyche can use various data sources to a) not only understand the precise make-up of the staff e.g. whether they are radiographers or GPs, two very different risks but also b) automatically create a risk model based on this data.
About the company
Tyche was founded in 2015 by Adam Cassady and David DeAngelis. The idea came about after Adam, an attorney at the time, was staffed on cases relating to Diacetyl. Diacetyl is a flavouring used in microwave popcorn with alleged links to lung disease and death in factory workers. According to Adam, despite a 2003 research showing this link, by 2006 insurers still didn’t have Diacetyl exclusions or conditions in policies. Adam Cassady, CEO: “if insurers had had better access to insight about this chemical at worst they could have saved $1Bn in exposure, at best saved lives. We asked ourselves how we could give insurers better visibility of external data. The answer was Tyche”.
Tyche are understandably coy about their client base. They claim now to have more than one major reinsurer and 3-10 mid-size casualty carriers in the US. Adam again: “We’re very pleased with progress we’ve made and the client interest. We are very busy”.
The Oxbow Partners view
Anything that improves underwriting results surely must be beneficial for industry and insured alike. We can, however, see two potential issues for companies like Tyche:
- Are insurers comfortable using underwriting tools where it is not possible to understand every step in the logic? Insurers and regulators like audit trails. AI is figuratively a black box.
- How easy is it for competitors to step in, analyse their own training data and copy the business model?
We also see a broader issue. At what point does underwriting become so accurate that you stop pooling the risk and, in essence, premiums just become a prepayment for future claims. Most likely, by that time the robots will have taken over, giving us all bigger concerns than the cost of insurance.