Curious what’s actually useful vs hype.

Wanted to get a real-world temperature check on this. My company in India (B2B SaaS) is being pushed by management to “use AI in our QA process” — classic top-down pressure without much direction.

I’ve personally tried:

  • GitHub Copilot for autocompleting test cases (actually useful for boilerplate)
  • Cursor for refactoring old Selenium tests (surprisingly decent)
  • One of the newer “autonomous testing” tools I won’t name ! complete waste of money

What’s been your experience? I’m specifically interested in whether anyone is using AI for: Test case generation from requirements documents

That seems like the most promising use case to me but I haven’t found a solid workflow yet.

  • perangkat_lunakEnglish
    0·
    2 months ago

    We use an LLM to convert Gherkin specs into Playwright test scaffolding. It’s not magic but it cuts the initial draft time by maybe 60%. A tester still reviews and completes every test.

    The key insight for us was: AI is good at structure and boilerplate 🦾 🦾 , humans are still needed for assertions and edge case logic. 💪💪

    Given / When / Then → AI scaffold → Human review → Done

    That pipeline actually works in practice. 💪