Author(s): Pranta Kundu

Originally published on Towards AI.

A Staff SDET’s field guide to the architectural shift — for juniors who want to actually understand it, not just hear the buzzword.

Your test fails at 2 AM. The pipeline goes red. A human wakes up, opens the trace viewer, stares at a screenshot, greps the logs, guesses it’s a selector change, fixes it, reruns, goes back to sleep.

After the lead, the article explains how agentic QA changes the automation loop from a predetermined “script” into a goal-driven “agent” that can reason, use tools, observe results, maintain state, and decide what to do next—while emphasizing that the underlying automation framework remains important. It breaks down what traditional frameworks do, what new capabilities (and risks) agents introduce, and why “self-healing” and “agentic” are not the same thing, including a comparison table. Using a concrete login-failure example, it contrasts traditional automation, AI-assisted (suggestions only), and agentic workflows (tool-backed actions like updating locators via draft PRs, rerunning tests, and escalating when unsure). The piece then outlines practical architecture patterns (tools vs. vision, observe/evaluate/guardrails, controlled reversible actions), identifies what agents still can’t reliably judge (business intent, acceptable risk, product judgment, masking defects), and provides guidance on what SDETs should learn next. It closes with a “what not to automate” no-go list, a readiness checklist for piloting agentic QA, and a maturity path showing the progression from AI-assisted to agentic test engineering and ultimately AI QA agent engineering.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI