MaddieJeanNielsen

A twenty-year-old consulting firm had never had a quality function - no test cases, no process, no methodology. I was hired as its first quality analyst and spent six years building the department: defining what a test case was, selecting automation tooling by project type, restructuring sprint cadence so quality wasn't last in line, and staffing a team across Chicago and Latin America. I productized it into a service the firm sold to clients. By the time I left it was one of the highest-margin teams in the company, and two analysts I'd developed had moved into senior roles.

The hardest part was never the testing. It was getting a company that had gone twenty years without the function to value the discipline, and you don't win that with a process document. I learned to sell the work before building it, to frame quality as something that made other people's jobs easier rather than a gate in front of them. Every version of that job, internal or client-facing, came down to the same thing. I've stopped believing a good idea carries itself.

I wanted a real site, not a template, hosted on my own terms, light enough to load fast, structured so AI search could actually read it, accessible, and editable without a deployment. It also had to be repeatable on free tiers, because the next one is for a community group with no budget. I wrote those constraints first, had AI evaluate options against them, and chose Astro and Sanity - an editor that ships raw HTML costs far less bandwidth than a full CMS. The design, the content model, the branching strategy, and every commit were mine. AI wrote the code. I ran the project.

The build was the easy part. The hard part was deciding what the thing had to do before I let anything get made - free tiers, no deploy to publish, light enough to load anywhere, readable by AI search. Once those were written down, the stack chose itself and I could tell whether what AI produced was right. Handing a model a vague ask gets you a vague result, which is the same problem I spent twelve years catching downstream.

I was never allowed to hire anyone at my level. Budget capped one at entry-level and the other at standard QA; so I got someone from manufacturing QA with no software experience and a game tester who'd never touched a web app, training both while shipping the MVP of our largest client project. Later hires came via an offshore acquisition, and I mentored our first automation engineer on strategy: what was worth automating and why. One made senior in a year, then lead; another made senior. Two earned ISTQB certification, and we were the first team in the company to integrate Chicago and LATAM.

I wanted a peer and never got one. It turned out to be the most useful constraint I've had - if nobody arrives senior, you have to build the thing that makes people senior: training material, a clear bar, and enough of your own time to actually spend on it. I learned to manage because there was no other way to get the team I needed. The promotions still mean more to me than anything I shipped.