Seven hundred applications arrive

Matt Stauffer, a forty-one-year-old CEO of Tighten, a programming company based in Atlanta, posted a project manager job on August twenty-ninth. The company is fully remote with around fifteen full-time team members. Stauffer received seven hundred three applications.

Why personal questions matter now

Stauffer included personal questions in the application to understand applicants' humanity and communication style. Questions asked how they pronounce GIF, their favorite movie, thoughts on the Oxford comma, and their most formative experience becoming a project manager.

AI patterns emerge quickly

By application thirty, Stauffer could tell some were fully AI-written. He studied English in school, writes blogs, and wrote a book, so he could recognize the patterns. The most obvious sign was the movie question.

The favorite movie giveaway

AI consistently suggested the same movies: The Pursuit of Happiness, Inception, The Secret Life of Pets, and The Lion King. Applicants using AI would write ten sentences explaining why Inception appealed to them as a project manager, which felt tonally wrong for the question.

“I really love Inception, and then the applicant went on to write 10 sentences explaining that they like Inception because they're a project manager and project managers like movies like Inception.”

Boilerplate instructions left behind

Some applicants accidentally copied AI's boilerplate instructions into their answers. Responses would begin with phrases like, Here's a great answer you should put that will show how great of a project manager you are, revealing the AI's prompt.

“Here's a great answer you should put that will show how great of a project manager you are.”

The Oxford comma pattern

For the Oxford comma question, AI answers consistently said reduces ambiguity in lists. After reading a few applications, Stauffer recognized this phrase as a reliable indicator of AI writing, even though AI models are non-deterministic.

“reduces ambiguity in lists”

AI is a tool, but use it right

Stauffer uses AI himself as a reviewer and conversation partner in writing and coding. He is not against AI, but worries that if he can spot AI writing, clients can too. The line is whether humans clearly shaped the final output.

“if I can tell you used AI, clients can too.”

Human involvement makes the difference

Stauffer accepts applications showing AI use if a human clearly prompted, reviewed, and edited their own humanity into the answer. He wants his team capable of writing if AI tools fail. He acknowledges he might be wrong about some applications.