Agentic Steel Thread for Test Automation

Blessan Thomas from Metro Bank presents the 'agentic steel thread', an AI-driven test automation framework designed to take a single test requirement through an entire end-to-end execution lifecycle. The presenter explains how an AI orchestrator utilizes a series of specialized skills—including export, design reading (e.g., via Figma), application knowledge building, test data management, and execution—to transform high-level, vague test cases into detailed, executable UI steps. The demonstration shows the framework handling complex scenarios such as flaky tests by looping back to a re-execution stage, managing test data via JSON, and integrating with tools like Jira X-ray and Perfecto AI. The session concludes with a discussion on the trade-offs of AI automation, such as token costs and the risk of reduced critical thinking, while highlighting the benefits of reduced maintenance and self-healing capabilities.