Revolutionizing QA With AI
Man, that would have been great if it worked (Spoiler: We’re Writing Cypress Tests Instead)
Dream
Starting a new job is humbling. New codebase, new stack, new product, and, in so many cases, QA practices that are*… in development*.
So when I built my first user-facing feature at this new gig, I wanted to QA it thoroughly. Not just “poke around until it seems fine” QA, but a real, reusable flow the team could rely on.
And then I had my big idea.
What if a developer could record themselves demoing the app, feed that transcript into AI, and instantly get a beautifully formatted QA checklist? Something we could standardize and reuse for UAT, regression, and feature testing.
The dream workflow looked like this:
Record a demo of the app (Zoom, Gemini, Loom, whatever).
Generate a transcript.
AI maps the transcript into a template (Confluence page? Gdocs? Excel sheet? CSV??).
Team has a living checklist, a repeatable QA workflow, without reinventing the wheel each time.
In my head, I was building the bridge between AI magic and practical QA.
Meet, reality
In practice? I was just building a headache.
CSVs were too rigid to capture nuanced flows.
Excel and Confluence were friendlier but bloated and impossible to maintain by hand, let alone with tools like Gemini or ChatGPT.
Centralizing all this into a flexible “database” that could handle multiple personas, full-application tours, and one-off feature demos? Not happening.
Somewhere along the way, I realized my “genius” workflow had quietly morphed into a second full-time job.
And here’s the reflective part: my heart was in the right place. I really wanted to give my team a process that scaled, that turned tacit knowledge into structured QA steps. But I overshot. I tried to solve everything with one clever system instead of solving the thing right in front of me.
What the team actually needed wasn’t a shiny AI workflow. They needed good Cypress tests: automated, reliable, boring in the best way possible.
So that’s where I landed. My “AI-powered QA revolution” is going back into the drawer of Overengineered Ideas™, and I’m refocusing on what will genuinely help: growing a strong, maintainable test suite.
Because the truth is, AI might help with QA someday. But today, the best “AI-assisted workflow” is still just me, writing the test, and hitting npm run cypress:open.
Is cy.prompt out yet?
