The Problem
We were hiring a new paid media strategist. We got 87 applications. Reviewing them took hours. Reading resumes, checking portfolios, comparing candidates. Managing it was a nightmare.
We needed to get to a top 5 quickly. So we asked: What if an agent reviewed these?
What We Built
An agent that:
1. Read each resume
2. Scored against our criteria (relevant experience, specific skills, portfolio quality)
3. Extracted key information (years of experience, relevant platforms, past clients)
4. Created a one-page summary of each candidate
5. Ranked the top candidates
6. Flagged concerns (resume gaps, missing portfolio, etc.)
Not to decide who to hire. Just to do the initial screening so our team could focus on the candidates that mattered.
What Happened
The agent screened all 87 in about 2 hours. Ranked them. Surfaced the top 10. Our team reviewed those 10 more carefully and picked 5 for interviews.
The agent's ranking was pretty good. Not perfect, but better than random.
We interviewed the top 5. One candidate was clearly the best fit. We made an offer. She accepted.
The Part That Matters
The second-highest ranked candidate on the agent's list? We might have skipped her if we were relying on gut feel. But the agent flagged her with a note: "Smaller portfolio but exact platform experience. Worth a conversation."
We interviewed her. She was solid. Not quite as good as our hire, but close.
The key insight: The agent's ranking didn't matter. The filtering and summarization mattered. It let us focus our energy on candidates who actually fit, instead of wading through 87 resumes.
The Red Flag Catch
Our top candidate initially seemed perfect. But the agent's screening flagged something we missed: Her resume claimed she ran campaigns for a competitor. But when we checked, that timeline didn't match her stated employment history.
Turns out it was a mistake on her resume. She explained it in the interview. But the agent caught the inconsistency before we wasted time in later-stage interviews with someone who had red flags.
What We Wouldn't Let the Agent Do
We didn't let the agent make the final decision. "Hire this person." That's human decision-making territory.
We didn't let it score candidates on "culture fit" because that's subjective and easy to bias.
We didn't automate away the interviews. That's where you actually learn about someone.
The Pattern
AI works in hiring when you're doing the grunt work filtering. "Sort these by relevance." "Flag inconsistencies." "Summarize key qualifications." Those are good uses.
AI doesn't work when you're doing judgment work. "Is this person going to be good at the job?" "Will they fit our culture?" "Can they handle the pressure?" That's interview territory.
The Time Savings
Screening 87 resumes by hand would have taken 8-10 hours for a hiring manager. The agent did it in 2 hours. The hiring manager spent 3 hours reviewing the agent's summaries and recommendations.
Net: 5-7 hours saved. Not life-changing. But enough to matter when you're hiring.
And the fact that we caught that resume inconsistency earlier? That might have saved us from a bad hire. Can't put a number on that.