Intuit Uses Agentic AI to Automate Multi-Region Disaster Recovery

Intuit has developed an agentic AI assistant to manage large-scale disaster recovery (DR) operations across its cloud infrastructure. The system automates the coordination of failovers for thousands of microservices across multiple AWS Regions, ensuring critical services remain available during regional degradations.

Built on Amazon Bedrock, the assistant employs an architectural pattern consisting of typed skills, a thin Bedrock layer, and a bounded agentic loop. This framework allows the AI to orchestrate the movement of assets from degraded primary regions to healthy secondary regions. To execute these changes, the assistant integrates with EWOK, Intuit's internal recovery execution layer, which handles the technical failover of assets. According to Amazon Web Services, the architecture is flexible enough to work with any system that exposes authenticated and auditable APIs, extending its utility beyond the EWOK layer.

The Scale of Recovery

Intuit manages a massive infrastructure footprint that powers high-traffic financial products, including TurboTax, QuickBooks, Mailchimp, and Credit Karma. In a multi-region AWS environment, coordinating a reliable failover across thousands of interdependent microservices is a significant operational hurdle. Traditionally, these processes relied on complex manual coordination or rigid scripting, both of which can be slow and prone to error when executed under the pressure of a live outage.

Shifting to Agentic Operations

This implementation marks a transition toward "agentic" operational tools. Unlike standard AI assistants that merely suggest a course of action, this system orchestrates complex, deterministic workflows through a bounded loop. By removing the need for manual intervention in the coordination phase, Intuit reduces the risk of human error during high-stress events. For the end user, this means a faster time to recovery for essential financial services and higher overall system resilience.

Future Applications

While currently focused on disaster recovery, the underlying design of the assistant suggests a broader application for autonomous infrastructure management. Because the system relies on a generalized pattern of typed skills and auditable APIs, Intuit can potentially adapt the agentic loop to other operational domains. The company's move signals a growing industry trend where AI is moving from the role of advisor to the role of orchestrator for mission-critical cloud operations.