Screen 50 Résumés in 30 Minutes Without Missing the Good Ones
A 6-step AI workflow that ranks an applicant pile against your actual scorecard, surfaces the top 8, and writes your first-round questions, with a human gate before anyone gets a no.
The problem: A job post pulls 80 applicants, you skim 15, hire from a shrunken pool, and tell yourself you ran a process. AI can read the whole pile against a consistent bar, but only if you give it a real scorecard first. Here's the workflow.
Step 1: Build the scorecard (you, 10 min). Before touching résumés, write 5–7 weighted criteria for this role: must-haves (e.g. "2+ yrs in field service scheduling"), nice-to-haves, and instant disqualifiers. Assign each a weight. This is the human judgment AI can't invent for you.
Step 2: Normalize the pile (ChatGPT or Claude). Export applications to a single doc or paste in batches. Prompt: "For each candidate, extract: name, years relevant experience, top 3 relevant skills, employment gaps >6 months, and any disqualifier from this list: [paste]." Get a clean table.
Step 3: Score against the card (AI). Prompt: "Score each candidate 0–100 using these weighted criteria: [paste scorecard]. Show the score, the 1-line reason, and flag anyone missing a must-have." Sort descending.
Step 4: Human gate (you, 10 min). 🚦 Pull the top ~12 and read those résumés yourself. AI ranks; you decide. Look for the things a model misses: trajectory, a side project, a red flag in how they wrote. Cut to your top 8.
Step 5: Generate tailored screen questions (AI). For each of the 8: "Write 3 phone-screen questions targeting this candidate's specific gaps or claims from their résumé." Now your 15-minute calls actually probe something.
Step 6: Close the loop kindly (AI-drafted, human-sent). Draft a warm, specific rejection template for the rest. Read before sending. Never let a model auto-send a no to a real person.
What it's worth: You go from reading 18% of your applicants to scoring 100% of them, with a consistent bar, in about half an hour. The best hire in the pile is often #34: the one you never opened.