AWS Builder Center
Revolutionizing Browser Automation with Amazon Nova Act and Amazon Bedrock AgentCore

Revolutionizing Browser Automation with Amazon Nova Act and Amazon Bedrock AgentCore

A practical guide to building intelligent browser automation agents using Amazon Nova Act and Amazon Bedrock AgentCore, bridging traditional RPA with modern agentic AI for complex, multi-step workflows in regulated industries.

Revolutionizing Browser Automation with Amazon Nova Act and Amazon Bedrock AgentCore



A leading enterprise automation platform leveraged Amazon Nova Act and Amazon Bedrock AgentCore to revolutionize browser automation capabilities, enabling sophisticated multi-step workflows and marketplace-ready intelligent agents. This implementation demonstrates how Nova Act bridges the gap between traditional Robotic Process Automation (RPA) and modern agentic AI, delivering production-ready solutions for complex verification workflows.



Business Challenge

The customer needed to enhance their automation platform with advanced browser-based AI capabilities that could:
  • Execute complex, multi-step browser workflows requiring sequential operations
  • Handle variable-length tasks with asynchronous processing
  • Integrate seamlessly with existing RPA infrastructure
  • Deliver marketplace-ready solutions for regulated industries
Traditional scripted automation struggled with dynamic web content, required frequent maintenance, and couldn't adapt to workflow variations without manual intervention. The customer sought a solution that combined the reliability of RPA with the flexibility of AI-powered agents.



Solution Architecture

DIY Browser Automation Framework

The customer implemented a flexible browser automation framework using Nova Act's three core actions:
Action
Purpose
Start Session
Initiates browser lifecycle management outside Nova agent scope
Run
Executes agentic operations with full context awareness
End Session
Cleanly terminates browser instances
Key Innovation: By managing the browser lifecycle externally, the framework enables sequential agent operations and allows users to chain multiple prompts while combining agentic actions with traditional RPA activities.

Technical Capabilities

  • Session naming for sophisticated window management
  • JSON output parsing for seamless data flow between agents
  • Flexible integration with existing automation workflows
  • Mixed automation support — combining agentic and traditional RPA within a single workflow



Primary Source Verification Agent (Marketplace Solution)

A production-ready agent built on AgentCore infrastructure that automates credential verification across multiple regulatory authority websites.

Solution Components

┌─────────────────────────────────────────────────────────┐
│ RESTful API Layer │
├─────────────────────────────────────────────────────────┤
│ Job Submission │ Status Polling │ Results Query │
└────────┬──────────┴────────┬─────────┴────────┬─────────┘
│ │ │
▼ ▼ ▼
┌─────────────────────────────────────────────────────────┐
│ AgentCore + Nova Act Engine │
│ • Asynchronous job processing (5-10 min execution) │
│ • Input processing (license numbers, types, URLs) │
│ • Automated navigation & data extraction │
└────────────────────────┬────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ Amazon DynamoDB (Results Store) │
│ • Structured results retrieval │
│ • Job status tracking │
└─────────────────────────────────────────────────────────┘
Business Impact: Eliminates manual verification processes, reduces processing time from hours to minutes, and ensures compliance with credentialing requirements in regulated industries.



Technical Implementation

Architecture Highlights

Component
Role
Amazon Bedrock AgentCore
Scalable, managed execution environment with isolated browser sessions
Decoupled Browser Lifecycle
Enables complex multi-step automations beyond single-prompt interactions
Structured Data Output
JSON-formatted results for downstream processing and integration
Amazon DynamoDB
Persistent storage for reliable, queryable data retrieval

Addressing Production Requirements

Authentication
  • Working toward SSO integration for internal applications
  • Secure credential management for external sites
  • Token-based session continuity across multi-step workflows
Scalability
  • Evaluating cost models for 24/7 workloads
  • Comparing AgentCore browser services against customer-owned VM infrastructure
  • Asynchronous processing to handle variable execution durations efficiently
Observability
  • Execution logging for troubleshooting and compliance
  • Exploring video recording capabilities for audit trails and debugging
  • CloudWatch integration for operational metrics



Results & Benefits

Operational Improvements

Metric
Improvement
Automation Velocity
Reduced development time for browser-based automations
Flexibility
Seamless integration of AI agents with existing RPA workflows
Scalability
Asynchronous processing handles variable execution times efficiently
Accuracy
AI-powered navigation adapts to UI changes without script maintenance

Strategic Advantages

  • Marketplace Readiness: Production-grade agents deployable to customer environments
  • Minimal Configuration: Human-like agents requiring less extensive setup compared to traditional scripted automation
  • Continuous Learning: Runtime feedback mechanisms for ongoing improvement
  • Reduced Maintenance: AI agents adapt to website UI changes that would break traditional selectors



Conclusion

This implementation demonstrates the transformative potential of combining traditional RPA with modern agentic AI. By leveraging Amazon Nova Act's browser automation capabilities and Amazon Bedrock AgentCore infrastructure, the customer created both flexible DIY frameworks and production-ready marketplace solutions that deliver measurable business value — while paving the way for more sophisticated, human-like automation agents.
The combination of Nova Act's intelligent browser interaction and AgentCore's managed infrastructure provides a compelling path for enterprises looking to evolve beyond traditional scripted automation toward adaptive, AI-powered workflows that can handle the complexity of real-world business processes.



Getting Started

To explore Amazon Nova Act and Amazon Bedrock AgentCore for your browser automation use cases:
  1.  Amazon Nova Act Documentation  — Learn about Nova Act's browser automation capabilities
  1.  Amazon Bedrock AgentCore  — Explore the managed runtime for AI agents
  1.  AWS Machine Learning Blog  — Agentic QA automation using AgentCore Browser and Nova Act


Published: August 2026
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article