Chubb has been working with Blaze Advisor to automate a number of decisions. They began with specialty lines underwriting (automated renewals), claims severity calculation and work queue assignment. Current focus is on integrating predictive models and some legacy modernization.
The automated renewals project reduced the time to make renewal rule changes from 3-6 months of IT work to 2-3 days of IT work. The project contains a rule maintenance application and a batch cycle that runs the renewals through the rule server based on the current set of rules in the repository. The rule maintenance application was based around what they called “levers” – each lever is an attribute that can be used as part of the renewal decision. There is also a What-If feature. This allows business users to see the results of a proposed change in the rules using a year’s worth of data. The project had no quality problems, generated some good teamwork and hit its schedule. The What-If capability, a decision simulator really, seems like a crucial component.
Disappointing lack of information about what Chubb has been doing and too much generic Fair Isaac stuff but oh well.