# Marcus Thibodeau, QA Lead at Fieldwork Software — read of QA Testing AI, May 14 2026

> 9 years in QA, currently holding the line for a 38-person SaaS team that ships bi-weekly. I coach U10 soccer Saturday mornings, which means I do most of my tool research on the 6:12 AM train before anyone else is awake.

## How I got here

Searching "generate test cases from user stories AI" on Google, which I've typed into a search bar at least six times this year. A Medium post about AI-assisted QA linked to this page. I had maybe four minutes before my stop. I kept reading past my stop. That's a thing that happens to me maybe once a month.

## What I clicked first

The hero line: "Your test suite catches maybe 60% of what breaks in production. QA Testing AI identifies the 40% your tests miss."

That specific number landed differently than the usual "ship faster" stuff. 60% coverage is embarrassingly close to where we actually are. Whether it's true or a guess, whoever wrote it has been in a QA standup before. Then: "Reduce QA cycle time by 70 percent." Okay, we're back to the usual.

## Where I paused

The "Results from Real Teams" section. Not because I believed it, but because it was specific in the right ways. "160 hours to 40 hours per release cycle" and "Compliance audit time dropped from 3 weeks to 1 week" are the kinds of numbers that come from someone who has actually watched a QA team work, not from someone who read a G2 category page. I paused because I wanted to find a name, a person, a company. There wasn't one.

And then I hit the bottom of the page. "Honest disclosure: we don't have live customers on this idea yet. We shipped the strategy package; you ship the customer conversations." I read that twice. The entire page above it is written as if the product exists. That is a strange experience.

## What I distrusted

The case study framing. "Series B SaaS (12 engineers, B2B operations)" is not a case study. It is a demographic sketch attached to some numbers. No quote. No name. No before-and-after story. It reads like someone asked an LLM to write plausible ROI stats and then gave them a thin label to make them feel sourced.

Also: "85 percent of regression testing automated" for the fintech platform. Automated by this tool specifically, or by the team in general over time? That distinction matters a lot to me and the page papers over it completely.

The "1 in 6 Meaningful-success odds (Fermi)" scored on the bottom of the page. That is not a thing you put on a product page for users. That is a thing you put on a page for people evaluating whether to BUILD this product. Which is what this actually is.

## What would convince me

One real company name and one QA engineer I can email. Not a quote on the page. An actual "reach out to Sarah at Flockjay, she'll tell you what happened" moment. That is the only evidence that moves me on a product that is making claims this specific.

If there were a short screen recording showing actual generated output for a real feature spec, not a toy example, that would do more than any stat on the page. Show me the Cypress output for a login flow with edge cases. Show me if it handles the "user has MFA enabled but loses their authenticator" branch. That is the test case nobody writes.

## What I'd ask in an email reply

1. The page says tests export as "production-ready" Cypress, Playwright, Selenium. What does that actually mean? Do the selectors hold up against a real DOM, or do I need a human to review every generated test before it goes in the pipeline?

2. When a UI change breaks 40 generated selectors at once, what does the maintenance workflow look like? That is the actual pain point and the page does not answer it.

3. The bottom of the page says this is a "strategy package" and there are no live customers. So what am I trialing if I start a free trial? Is there an actual tool running, or is this a waitlist dressed up as a product page?

## Verdict: on-the-fence

The page understands my problems better than most. But it is selling a product that does not appear to exist yet, using case studies that have no names attached to them, which means the 9 years of scar tissue I have around "this tool will change everything" is fully activated. If someone answered question three honestly I would probably reply.

---
*Memo by skeptic persona, generated 2026-05-14. Studio breaks own self-grading loop.*