Published in Product
How Tandem screens 200 applicants before lunch
Post a job for a decent role and the applications arrive faster than anyone can read them. Two hundred in the first morning is normal now — AI-written resumes have made applying nearly free, which means the cost has moved entirely to the reviewer. Most teams respond by skimming, keyword-matching, or quietly giving up on the bottom half of the pile.
Tandem's screening agent takes a different position: every applicant gets a real read, against your actual bar, within minutes of applying. Here's how that works under the hood, and why doing it fast isn't a shortcut — it's what makes the screening honest.
The pile is the problem
Human screening degrades with volume. The first twenty resumes get careful attention; by the eightieth, reviewers are pattern-matching on school names and job titles. Studies of recruiter behavior put the median resume read at under ten seconds — which isn't reading, it's triage. The candidates who lose in that regime aren't the weakest ones. They're the ones with unconventional formats, non-linear careers, or resumes that don't front-load the right nouns.
An agent doesn't fatigue. The 200th application gets exactly the same treatment as the first: full parse, evidence extraction, scoring against the same rubric. Consistency isn't a nice-to-have in screening — it's the whole point. A bar applied inconsistently isn't a bar; it's a mood.
Evidence, not keywords
Keyword filters ask "does the resume contain the word Kubernetes?" Tandem asks "what has this person actually done, and does it support the claim?" The screening agent reads each application the way a strong hiring manager would: it extracts concrete evidence — projects shipped, scope owned, outcomes described — and maps each piece to the requirements in your hiring plan.
That distinction matters most at the margins. A candidate who wrote "migrated a monolith serving 2M users to services" but never typed the word "microservices" passes an evidence read and fails a keyword filter. A candidate who pasted the job description into their skills section does the opposite. Screening on evidence flips who survives — usually in your favor.
Your bar is the rubric
The agent doesn't invent standards. When you build a role in Tandem, the hiring plan captures your must-haves and nice-to-haves in plain language — "has owned a production service end to end," not "5+ years experience." That plan becomes the rubric the agent scores against, verbatim. If a requirement is ambiguous, the agent flags it during planning rather than guessing at screen time.
Every screening decision comes back with receipts: which requirements the candidate met, the specific evidence for each, and what's missing. When you disagree with a call, you correct the plan — not the individual decision — and the correction applies to every applicant after that. The bar gets sharper as you use it.
Why speed changes quality
Screening 200 applicants before lunch sounds like a throughput brag, but the real effect is on the candidates. The best people in your pile are interviewing elsewhere this week. When screening takes ten days, your top candidates hear from you after they've signed somewhere else — so slow screening doesn't just delay your pipeline, it silently filters out the people with the most options.
Fast, same-morning screening means a strong applicant gets an interview invitation while they're still deciding where to focus. Teams on Tandem routinely book first interviews within a day of an application arriving. The pile stops being a backlog and becomes what it should have been all along: a live stream of decisions, each one made while it still matters.