Libra – First Autonomous AI product stack for insurance

Claims settlement API Policy issuance Product recommendation Advisory Planning Automated Claims Processing (ACP) END SOLUTIONS Autonomous Underwriting Module (AUM) Compliance and Risk Monitoring (CARM) Assistive AI (AAI) TOOL SET SHARED LEARNING Model building Model optimization Simulation AI LIBRA MODEL MAINTENANCE DEPLOYMENT ENVIRONMENT Sales data SECURITY Cloud On-Premise Hybrid Renewals data Marketing data Third Party Claims Data THIRD PARTY DATA USER TOUCH POINTS DATA PIPELINES

Holistic fully connected platform

The path to implementing a full-scale autonomous AI driven enterprise may be long and daunting, hence Arya’s architecture makes it possible to take one step at a time. Arya’s ‘Libra’ is an Insurance product stack that is specific to the insurance industry and is broad enough to automate or assist all key processes.

Enterprises can begin with just one module on the platform. Once this stabilizes and gains organizational acceptance, the next module can be installed. The best part is that the efforts and intelligence of the first module plays into the next module being implemented.

Plug and usable products & modules

Modular nature of the stack allows insurers to scale progressively and deploy products as per priority. This delivers much needed speed and flexibility in customising these product and at the same time to reuse the efforts and optimizing it for next incremental deployments. Such products are – Autonomous Underwriting Module, Automated Claims Processing, Compliance and Risk Monitoring and Assistive AI.

Products-on-a-platform: Shared learnings

The platform plus product strategy originated from our objective of automating all the business processes being driven in an Insurance company. The platform Libra enables intelligence sharing and cumulatively, individual solutions, all strengthen each other.

Put simply, the products on the platform, can share intelligence. The platform also enables models to be versioned, and stores model performance data for ongoing enhancements. An Arya’s product/module that is implemented today for Claims would be able to share intelligence and enhance the Underwriting module implemented tomorrow.

Faster training and customization

Being data driven products, the speed at which these can be trained determine how fast they can be deployed, and while in production how swiftly they can be customized. The platform enables rapid (re)training wherein data preparation, multi cycle experimentation, back testing can all be performed with minimal effort. Thus the decision chain from whether retraining is required to whether an updated module version is mature enough to be deployed is automated by the platform, with human approvals weaved in. 

Automated module scaling and versioning, makes it easy to maintain

Libra leverages VEGA’s platform functions like GUI framework, Simulation AI, Transfer learning etc. and offers the same deliverables within the product stack. These functions allows users to automate the complex data science tasks like hyper-parameter optimization, compute scaling and descaling, model versioning etc.

Once deployed, Libra monitors module performance and scales deployment to meet trending or peak seasonal loads. This involves adjustments, in tandem, of module serving architecture including APIs and also computing resources.

Tools to enable the enterprise to control the pace and direction of AI adoption

From GUI frameworks to Simulation AI, these data science tools on VEGA allows the insurers to use the same stack in building new deep learning modules. With VEGA, a user can focus on the essential task of building an AI model, without wasting time or attention on the sidelines for software and hardware dependencies and compatibility issues.

Starting your AI journey?

Leverage our team experience in deep learning and autonomous systems in financial services to productionize complex AI system in production through our off the shelves products or platforms.

Ideate, Innovate & Stay ahead!

Connect with arya.ai