Insurance is an $8 trillion global industry facing manual workflows and a growing talent shortage. Cara offers a native AI solution on AWS that automates back-office processes for insurance brokers.
Insurance agents regularly spend hours on repetitive tasks. This includes completing applications, analyzing insurance coverages, re-entering data between systems, and relaying information between customers and carriers. As the industry faces a persistent talent shortage, brokerages must increase revenue without proportionally increasing their workforce.
In this article, we explore how Cara, built in cooperation with AWS, addresses these challenges. We review the technical design decisions and AWS services that support the solution. We also share the measurable results Cara has achieved for corporate brokerages.
The challenge: why generic AI fails to meet insurance expectations
Insurance brokerages operate in a highly regulated environment. Every transaction requires accuracy, auditability and compliance. The data involved includes personally identifiable information (PII), financial records and subscription details.
Generic AI tools are not designed for this complexity. Effective AI for insurance must understand domain-specific data models and brokerage workflows. It must also meet the specific requirements of operators and regulatory constraints while respecting the company’s safety standards.
Cara’s founding team witnessed these shortcomings firsthand. Vic Yeh, Nikhil Kansal and Jon Patel previously founded a digital insurance brokerage. They developed it and sold it to The McGowan Companies, one of the largest private insurance organizations in the United States.
During this experiment, the team built an internal AI co-pilot powered by large language models (LLM). Copilot reduced turnaround times, improved data accuracy, and streamlined agent workflows. Encouraged by strong adoption, they expanded the concept into a standalone product: Cara.
Architecture overview
Cara is built on AWS services chosen for their reliability, scalability and security. Figure 1 shows the high-level components of Cara’s production deployment.
Cara architecture on AWS
Calculation and orchestration
Cara runs on Amazon Elastic Kubernetes Service (Amazon EKS) for container orchestration across multiple Availability Zones. EKS manages Cara’s microservices, including ingestion pipelines, workflow engines, and the inference layer.
This architecture supports elastic scaling to manage demand during peak renewal and maintenance periods. It supports thousands of concurrent users and workflows per brokerage. Each organization’s workloads run in isolated namespaces for tenant separation.
AI and inference
Cara’s AI capabilities are powered by LLMs hosted on Amazon Bedrock. Amazon Bedrock provides access to base models through a fully managed API. This allows Cara to run inferences without managing GPU infrastructure. Cara uses Amazon Bedrock for several main features:
- Coverage and quote information – compares carrier quotes, summarizes coverage differences, and highlights exclusions or gaps.
- Automation of applications and forms – completes ACORD and supplemental forms using source documents, previous submissions and agency guidelines.
- Generating proposals and renewals – produces client-ready brand renewal proposals and spreadsheets.
- Knowledge-based workflows – references agency-specific guidelines, carrier appetites and historical placements to guide decisions.
Security and data isolation
Data protection is a fundamental requirement for insurance organizations. Cara’s architecture uses account-specific deployments on AWS. Each brokerage’s data and workflows are isolated in dedicated, secure workspaces. This design supports compliance with industry regulations and provides organization-level auditability.
Integrations
Cara integrates with leading Agency Management Systems (AMS) and Customer Relationship Management (CRM) tools. It synchronizes accounts, policies and documents to reduce duplicate data entry. AI-powered workflows work directly within existing broker technology stacks. This design helps minimize changes to the systems their agents already use.
Deployment and operational characteristics
One of Cara’s design goals is rapid payback. Enterprise brokerages can be onboarded in hours and launch custom workflows in days. The Cara deployment on EKS uses templates configured for each new tenant. It provides isolated namespaces, storage and inference endpoints without manual configuration.
In production, Cara’s infrastructure on AWS provides:
- High availability – Multi-AZ deployment on EKS with automated failover.
- Elastic scaling – Kubernetes Horizontal Pod Autoscaler adjusts capacity based on real-time demand. This supports thousands of concurrent users during peak periods.
- Business Security – tenant-based data isolation, encryption at rest and in transit, and integration with AWS Identity and Access Management (AWS IAM).
Measurable results
Cara’s AI-powered workflows have delivered quantifiable results for insurance brokerages:
| Metric | Result |
| Time saved per user | ~10 hours per week through workflow automation and contextual knowledge retrieval |
| Integration speed | Corporate brokerages integrated within hours; custom workflows are up and running within days |
| Concurrent Capacity | Thousands of concurrent users and workflows per brokerage |
| Adoption | Used by hundreds of major insurance agencies and brokers |
These results come from automating organization-specific workflows and retrieving contextual knowledge. They rely on Cara’s domain-specific AI and the scalable, secure infrastructure provided by AWS.
Looking to the future
The insurance industry is still in the early stages of AI adoption. As business demand increases, Cara continues to expand its AI-powered workflows across sales, service and operations.
“We are excited to push the boundaries of domain-specific AI in real-world insurance use cases with AWS,” said Vic Yeh, CEO of Cara. “Our goal is to help insurance professionals get back to the heart of our industry: relationships.”
Conclusion
In this article, we showed how Cara created a domain-specific AI solution for insurance brokers using Amazon EKS and Amazon Bedrock. The architecture provides isolated and elastically scalable workspaces. It supports thousands of concurrent users while meeting the security and compliance requirements of the insurance industry.
To learn more about building AI-powered applications on AWS, visit the AWS Architecture Center. To get started with Amazon Bedrock, see Getting started with Amazon Bedrock. For Amazon EKS, see Getting started with Amazon EKS.
About the authors
Praise to Babul
Amaan is an Associate Solutions Architect at Amazon Web Services on the Startups team, based in Austin, Texas. He is passionate about helping startups build scalable, well-architected solutions on AWS, with a focus on AI/ML, generative AI, and modern application development.
For more information, please visit the original source Here.
“`
