AI in QA

Where AI helps NOW

• Test case generation — from user story + AI suggest.
• Visual regression smart diff — Applitools Eyes.
• Self-healing tests — locator broken → AI suggest fix.
• Bug clustering — Crashlytics ML group similar crash.
• Test prioritization — predict which test most likely fail.
• Code completion — Copilot for test code.

Tools available

• Maestro AI — generate flow from prompt.
• mabl — AI-driven test platform.
• Functionize — AI test maintenance.
• TestRigor — natural language test.
• Sauce Labs Sauce AI — flaky test detection.

What AI WON'T do (yet)

• Replace exploratory testing intuition.
• Truly understand user empathy.
• Design test strategy.
• Lead incident response.
• Stakeholder communication.

Practical AI workflows for QA today

Generate test cases from spec

Prompt to Claude/GPT-4: > "Given this API spec, generate 10 test cases including positive, negative, edge case."

Bug report enhancement

QA writes raw report → AI structure into template.

Test code review

AI scan PR → flag missing test, weak assertion.

Documentation

• Auto-summarize daily test result.
• Generate runbook from incident transcript.

Risk areas

• AI-generated test → looks plausible but wrong.
• Over-rely → atrophy testing skill.
• Privacy: don't paste prod data in AI tool.
• Bias: AI may miss edge case for underrepresented users.

VN context

• FPT.AI testing assistant in development.
• Most VN team experiment AI 2024-2026, mainstream adoption 2027+.

Skill for next 5 years

• Prompt engineering for QA.
• AI tool evaluation.
• Pair with AI, not replace.