Cross River builds agentic AI foundation with Tiger Analytics on AWS

Advancing an enterprise AI transformation strategy, Cross River needed a scalable way to build, deploy, and govern AI agents across the organization. The regulated financial services institution supports fintech partners and banking services. And it wanted to empower teams to use agentic AI to improve productivity across functions while maintaining strong oversight, security, and control. The goal was to create a reusable foundation that could support agent development across risk, finance, human resources, compliance, reporting, and other banking workflows.

Cross River recognized that conventional automation couldn’t address the full complexity of these workflows. So the company needed a governed, intelligent orchestration platform to reduce fragmentation while meeting transparency and compliance standards. By establishing a scalable foundation for governed agentic AI, Cross River democratized AI development across the organization while preparing for broader production deployment.

Working with Tiger Analytics, the company built Cross River AI HQ by using AgentFlow, the partner’s accelerator for designing, deploying, and monitoring agentic AI systems. The solution uses Amazon Elastic Kubernetes Service (Amazon EKS) for building, running, and scaling production-ready Kubernetes applications. To define and launch AWS resources in a logically isolated virtual network, the team deployed the Amazon EKS cluster privately within Amazon Virtual Private Cloud (Amazon VPC). Automatically scaling on demand, containerized agent workloads that run on Amazon EKS handle compliance, document processing, and partner reporting. Amazon Bedrock gives access to hundreds of foundation models from leading AI companies. For example, Anthropic’s Claude Sonnet powers the large language model layer to support language tasks across agent workflows.

By building governed data flows into AgentFlow, Tiger Analytics enabled banking systems, compliance databases, and partner APIs to feed agent-ready workflows. For the ingestion and preprocessing pipelines, Cross River and Tiger Analytics adopted two AWS services. Using Amazon SageMaker AI, the solution provides an integrated experience for analytics and AI. And by implementing AWS Lambda, the team runs code without thinking about servers or clusters. Raw data and processed data land in Amazon Simple Storage Service (Amazon S3), object storage that’s built to retrieve virtually any amount of data from anywhere. To give operations and risk teams the data they need for regulatory audit readiness, Cross River uses Amazon CloudWatch, a service for observing and optimizing workloads at scale.

To support regulated banking workflows, the team embedded governance controls into the Cross River AI HQ architecture. The solution uses AWS Identity and Access Management (AWS IAM) to securely manage identities and access to AWS services and resources. This way, nonhuman-identity architecture can assign unique, just-in-time credentials to agents, tools, and connectors, reducing credential sprawl across the migrated workflows. Granular role-based access controls enforce least-privilege policies, and configurable guardrails support content safety, risk-tier controls, and output filtering. To support operational-risk requirements, the team incorporated human-in-the-loop approval gates across production deployments and write operations. Also, a tiered control framework governs workflow migration by risk level. Low-risk processes, such as research and document drafting, move first. High-risk workflows follow only after full governance controls have been validated.

Cross River reduced operational complexity by replacing fragmented, manual processes with unified, event-driven AI pipelines. With Cross River AI HQ on AWS and AgentFlow from Tiger Analytics, the company created a scalable foundation for modern banking infrastructure. The team designed the solution to support AI agents at scale. “Working with Tiger Analytics and using AWS, we’ve built a foundation that helps us rapidly develop and deploy agentic AI across banking functions,” says Pravesh Rijal, chief data and AI officer at Cross River. “Several agents are already in production, with more than 100 underway to support the scale and pace of our AI ambitions.”

Tiger Analytics provided Cross River with a repeatable model that includes nonhuman-identity controls, human approval gates, role-based access, near real-time observability, audit trails, and risk-tiered migration. So the company can now expand its operational autonomy by reusing these capabilities for additional use cases and workflows.

Tiger Analytics offers AI and analytics consulting and solutions that help enterprises accelerate transformation.