Allstate case study

Predictive triage at the edge of operations.

Multi-million-dollar annual savings by triaging which homes actually needed an inspector. The same agentic-routing pattern now extends to claims, dispatch, and any high-volume operational decision where an agent can act before a human ever has to.

openworld

Presented at Oracle OpenWorld 2013

  • banner

    Using predictive model for potential $2-3M annual savings

  • banner

    No negative impact to claim rates.

Details, Numbers, and Highlights

The Challenge

  • Allstate inspected all properties it insured (100%)—there was a high cost associated with this exercise
  • Disconnect between capability to predict what house should be inspected and the ability to determine which houses to inspect and connect to operational dispatching
  • No tracking at the inspector in terms of accuracy of inspection; customer impact as well in terms of wrong information delivered to customers
  • 1 million inspections annually

The Solution

  • Reduce the # inspections – our models + internal models to predict when an inspection should take place considering agent, property, and internal models
  • Model the attributes about the agent; close loop on claims vs agent assessment of risk; huge cost and revenue impact

The Results

  • Potential millions in annual savings by accurately predicting which homes to inspect using predictive modelling
  • Delivered a sustainable solution for ongoing use
  • Improve customer experience– don’t have to be home for a home inspection
  • Provided opportunity for improvement in cost to model inspection samples (and related cost savings)