"Please share your screen."
Four words that have become a signature in recent interviews. Watching how someone works is still the best way to understand if someone is qualified for a job. I'd argue that most people are watching to catch someone using AI instead of evaluating how they’re using it… and you know exactly which one you are.
We get it, the paranoia is fair. By one analysis, 9 out of 10 resumes now contain inconsistencies such as overlapping dates or impossible promotions, often thanks to sloppy AI use. The industry’s answer is to catch them faster. Detection tools improve every quarter, models improve every week, so congrats, your detection tool is permanently one version behind. And not to mention, half of what it flags is a human who likes em-dashes.
Imagine two candidates who both used AI on their resume. The first pasted the job description into their LLM of choice and asked it to rewrite everything to match. It worked, the resume looks great, except the person behind it has no idea what any of it means and nothing in the process will find that out. The second candidate uses AI every day to move faster and make better decisions, they bounce ideas off it, they know what to keep and what to trash. Their resume looks almost identical to the first one. The detection system can’t tell them apart.
Why punish someone for using AI? We don't punish people for using the internet. In fact, we would laugh if someone said, "you can't use the internet to solve this problem."
What if we did the opposite? Give candidates a task, let them use AI if they’d like, and review how they use it and what it does to their output. Work samples have been around for decades, but they were always time-consuming and expensive, so hiring teams built proxies like resume screens instead. Resumes are dead. Evaluate people based on what they can do and if they can do the job you’re hiring for effectively.
MeritFirst makes work samples easy to run by creating assessments that mirror what candidates would be tackling day to day. On a resume, the two candidates may look identical, but when showing how they would go about solving a problem, they may look vastly different.
Yes, "please share your screen" can still be the right opener, but not to catch someone using AI… it's to understand how they're using it. The output is either good or it isn't and how they got there is the part worth watching.
Easy to nod along. Harder to pass one. Try it here.