A writer who followed the words into the machine.

Most people never think about the words inside software until those words let them down. A button says the wrong thing. A help article answers a question nobody asked. An assistant greets a brand-new customer with “How can I help?” when they don’t yet know what to ask. I’ve spent my career on the other side of those moments.

I studied journalism at Georgia State, and I still work like a reporter. Before I write anything, I want to know who’s reading, what they already understand and what they need to do next. My early jobs were in communications, including running web and campaign content for Georgia Tech’s College of Design.

In 2018, I joined Mailchimp as a technical writer. Over the next eight years, I went from writing help articles to leading a team of more than ten content designers, content engineers, instructional designers, and video producers. Together we built a help center in six languages that reached more than 5 million people a year, and most customers who used it found their answer without contacting support.

Then the help center got a new reader: an AI assistant that answered customers from everything we’d written. I led the Mailchimp Assistant with a small team of conversation designers, and it changed how I thought about the work. A confusing sentence in an article became a confusing answer in a chat. A correct answer could still leave someone stuck. So we rewrote what “good” meant for marketers and small businesses, replaced walls of text with step-by-step help, and fixed the source content whenever the AI stumbled. In seven months, the share of conversations customers finished without needing a person went from 55% to 80%.

Today I work on AI evaluation as a contract content engineer at Meta. It’s the measurement side of the same question: what does a good answer look like, and how do you tell when one isn’t?