The insurance industry in general is cautious about the adoption of new technology, and the ingestion of risk data quite often begins on hard-copy paper or in an Excel spreadsheet. This data may be entered in practically any format whatsoever before being delivered, for example, to an insurance broker. The broker may then add additional information such as the insurance coverage and terms, before pushing the data out to underwriters in the market to quote. This process and the accompanying required data transformations, is a very manual, repetitive, and tedious one, especially due to the often liberal use of unstructured data and file types.
Pro Insurance Solutions Limited (“Pro”) provides an outsourced service where they receive broker data (either directly or via an insurer) and create a template (manually) to represent that risk. Pro would then greatly enhance that data with other available information, providing to customers a generalized template more readily ingestible by 3rd party models to price that risk through computer simulations in order to determine the theoretical cost of coverage. However, these simulators required fixed data formats, while Pro and their clients received the data in random formats.
Analycat Limited, a partner of Pro, is a UK-based AI and algorithm company with specialties in insurance as well as generalized AI.
To solve these process and data issues, Pro and Analycat decided to build a new “Cleansing Artificial Intelligence” (CAI) system that would automate these data transformation and data cleaning processes. The goal was to reduce the total end-to-end processing time from several days to several hours or less. This would give Pro’s clients a tremendous competitive advantage, and in fact, be a major market disruptor. Wintellect, in partnership with Pro and Analycat, delivered the system in late 2019.
The core vision of CAI was to create a system that continuously monitors Input Events (input of spreadsheet risk data) and then semi-automates, via workflow processes, the tasks of cleansing, transforming, and generating output documents. The application was first developed on-premise, and then later migrated to the Azure cloud using the App Service Migration tool. Once in production, monitoring and alerting capabilities were added using App Insights, and geo replication was used to provided failover and uptime guarantees.
The high-level conceptual architecture is as follows:
- Front end UI built using modern web technologies such as React and ASP.NET MVC.
- Hosting and services environments built using Microsoft Azure cloud technologies, including Azure AD, Azure Web Apps, and Application Insight to capture telemetry
- Data for the system was hosted and processed using a combination of storage technologies, including Azure Blob storage, Cosmos DB, and Azure Key Vault
- In addition Azure Cognitive Services was used for language Translation, and major use was made of Azure Search for fuzzy logic searches.
- SignalR was used to enable real-time functionality for the system
The high-level system functionality included the following:
Basic Document Flow Management (data and document ingest and output). This was driven by an Azure Logic app that triggered workflows based on the receipt of incoming emails. End-User Application Core Features were extensive, and included:
- Application Sign In to the Application Portal
- User Management
- Client Account Management
- Data Import (Triage Interface)
- Task Management
- Preliminary Analysis of Documents
- Document Cleansing
- Data Transformation
- Geocoding Interface
- Output Generation
- Data Management
- Schema Management
The core tools used consisted of the following:
- Power BI. Power BI is a suite of business analytics tools that can be used to
analyze data and share insights. Power BI reports can be connected to many different data sources, produce reports, and publish them for an organization to consume on the web and across mobile devices.
- Azure Blob Storage. Azure Blob storage supports storing unstructured data in the cloud and can be used to store any type of text or binary data.
- Azure Cosmos DB. Azure Cosmos DB is a globally distributed, multi-model database service that offers storage across several well-defined consistency levels.
- React. A framework used for creating browser applications.
- ASP.NET MVC. A framework for building web APIs that are used to support the front-end browser application.
- Also: Azure Search, Cognitive Services for Translation, and SignalR.
Azure DevOps for Application Lifecycle Management
To ensure the application could be rapidly built, iterated on, and deployed, Azure DevOps was used for application lifecycle management, automating the build and release cycle. The team implemented continuous integration and continuous delivery processes using Azure DevOps Pipelines, ensuring code could be built, tested, and released on demand. Finally, Azure ARM templates were used as the solution for infrastructure-as-code, ensuring the entire environment could be released to Azure in a fully automated manner.
Fig 3 - Transforming origin data into standard form using transformation formulas
Fig 4 - Capturing geocoding information for an individual property
Fig 5 - Mapping relationships between source columns and target columns
Figure 6 - Decision Tree Builder allows full configuration of processing logic
Pro and Analycat needed a partner that was expert at providing the architectural skills needed to take the complex business logic required and translate that into a technical design that was flexible and performant, as well as the implementation expertise to develop the production system.
- Wintellect has deep experience in developing intelligent and configurable decision solutions.
- Wintellect is a Microsoft Gold Cloud Platform, Application Development, Data Platform, and Data Analytics, partner
- Core focus and skillset in React for Modern Web Development.
- Wintellect is a recognized leader in software architecture and implementation on the web, mobile, and cloud platforms.