SmartIntake Project Final Report

Commercial Insurance’s processes are manual, complex, case specific in nature, and have comparatively lower data volumes vis-a-vis the fast flow and more standardised processes for Personal Insurance. These factors have historically made Commercial Insurance less suitable for applying Artificial Intelligence (AI) / Machine Learning (ML) technologies for productivity / efficiency improvements.

The development in deep learning models that are based on Transformer architectures and leverages transfer learning, however, have demonstrated great potential to breaking down these barriers, and achieving promising level of performance even with low level of use case specific training data.

A Proof-of-concept (POC) is commissioned, enabled by a grant awarded by the Monetary Authority of Singapore under the Financial Sector Technology & Innovation (FSTI) Scheme, with view to assess the feasibility to applying AI/ML models with transfer learning to facilitate the undertaking of a traditionally manual and complex Commercial Insurance process at the Singapore Branch of Zurich Insurance Company Limited (Zurich Singapore).

A Whitepaper has been prepared (see link below for download) to set out the background, use case under consideration, performance of results observed, as well as sharing with the readers the challenges and lessons learnt observed, and our vision and next steps following the POC.

We would like to thank all stakeholders and contributors to this project for their hard work and support to put Zurich Singapore in a good place to further embrace digital opportunities and better realise Zurich’s strategy to simplify our way of working, innovate, and deliver superior experience to our customers at an ever-faster pace.  We are also grateful for the feedback from the evaluation panel, who has taken time out to review the proposal of this POC, which has made this feasibility study possible.