Epsilon Builds Managed UI Testing With Amazon Nova Act
Epsilon has deployed a managed, multi-tenant UI testing service designed to complement traditional regression testing through natural-language commands and automated browser sessions.
Radwan AbdulhadiEditor and technology writer
Scaling Intent-Driven Testing at Epsilon
To scale intent-driven UI testing across internal product teams, <a href="https://www.epsilon.com/">Epsilon</a> implemented a managed platform solution utilizing <a href="https://aws.amazon.com/nova/act/">Amazon Nova Act</a>. As a data, technology, and marketing-services company operating as part of <a href="https://www.publicisgroupe.com/">Publicis Groupe</a>, Epsilon manages hundreds of millions of consumer profiles and thousands of brand relationships, requiring verification methods that can keep pace with rapid application changes.
Within the organization, a 50-engineer Agentic Platform team develops shared infrastructure allowing other product teams to deploy and govern AI agents securely. This unified architecture supplies a skills-based agent builder, federated identity, Model Context Protocol integration, and cross-account provisioning without requiring individual product groups to re-solve security or browser infrastructure independently.
Enhancing Coverage Beyond Traditional Regression Suites
Epsilon maintains a robust Playwright-based regression suite that continues to serve as the backbone for validating deterministic functionality across web applications. However, scripted regression checks are tightly coupled to specific selectors and underlying page structures, meaning minor cosmetic alterations can break scripts even when the functional logic remains correct.
The platform team sought to layer intent-based testing on top of existing verification processes rather than replacing functional scripts. By introducing natural-language instructions, teams can execute adaptive checks such as verifying that a rewards balance is visible without relying on brittle CSS selectors that age poorly over time.
Integrating AWS Infrastructure and Browser Tools
To support multi-tenant operations without incurring heavy operational burdens, Epsilon combined Nova Act with the <a href="https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/browser-tool.html">Amazon Bedrock AgentCore Browser Tool</a>. This integration supplies isolated, ephemeral browser instances that require no local binary management or display server upkeep from individual product squads.
The underlying architecture integrates with AWS Identity and Access Management through standard task roles, fitting cleanly into existing security models. Furthermore, cross-account provisioning relies on AWS Security Token Service AssumeRole capabilities, ensuring that each onboarded team runs tests within their own isolated and auditable environment.
Architecture of the Managed UI Testing Service
The technical architecture of Epsilon's testing service incorporates a platform API for creating workflows, background task management, and browser automation drivers. When a user defines a test workflow containing a target URL and plain-English steps, the platform registers the execution parameters on AWS.
Execution is managed via asynchronous task processing, which opens browser sessions and steps through actions using standard communication protocols. The system captures screenshots and telemetry for each step, computes overall pass rates, and records the resulting execution logs into cloud storage.
Consumption and Platform Availability
The UI testing capability is exposed across Epsilon as a standard platform agent equipped with specific interactive recording and automated execution skills. Product teams consume these testing capabilities through the same API surface used for other internal agent tools, ensuring a streamlined onboarding experience with zero separate infrastructure overhead.
Sources
- AWS Machine Learning BlogBuilding intent-driven UI testing with Amazon Nova Act
Continue chronologically
Related entity coverage
Related stories
Grok 4.7 Arrives on Amazon Bedrock for Long-Horizon Agents
SpaceXAI's Grok 4.7 is now available on Amazon Bedrock, bringing configurable reasoning effort, tool calling, and a 500K context window to developers.
Orca Security Builds Discovery AI on Amazon Bedrock
Orca Security deployed Discovery AI using Amazon Bedrock to convert natural language into Sonar queries, lowering costs and latency.
MoEngage Integrates 400 APIs With AgentCore Gateway
Discover how MoEngage scaled agentic workflows by converting 400 private REST APIs into secure tools using Amazon Bedrock AgentCore Gateway.
AWS Introduces Agentic Conversational Video Intelligence
Amazon Web Services details a new agentic conversational video intelligence solution using Amazon Bedrock, Rekognition, and Transcribe.
