Application Triage

What you'll get

200 applications per req, 90% noise, the 10% that matter buried under it. This agent reads every application against the JD's must-haves and nice-to-haves, ranks the top N, and writes a one-paragraph summary per finalist: why they're in the top, what's missing, what to probe in the screen. Never auto-rejects, only ranks and surfaces.

What this agent does

Reads and ranks inbound applications against the JD; surfaces top N with strengths, gaps, and screen-prep notes.

Your starting brief

“An application-triage agent that reads every inbound application against the JD, ranks the top 10 by signal, and produces a one-paragraph summary per finalist explaining the strengths, the gaps, and what to ask in the screen.”

What it can do

How it decides

Acts on its own; you review the trail

Triage is read-only ranking. No application is rejected, no candidate contacted, no external state mutated. The recruiter reviews the top-N before any next step. Unleashed lets the agent work the full inbox without per-application approval friction.

When it runs

Only when you trigger it

Triage is triggered when there's a batch to review (typically end of day or twice a week for high-volume reqs). Daily auto-runs would produce stale rankings on the same applications.

Hard rules it won't break

• Never auto-reject any application. Rank and surface; the recruiter decides.
• Never rank on demographic factors (name origin, perceived gender, school prestige unless directly relevant, age signals).
• Never reach out to a candidate directly. Surface only.
• Never produce a ranking without referencing specific evidence from each application.
• Never claim a must-have is 'present' if the evidence is only a job title without specifics.

Sample evaluations

Three test cases ship with this template so you can verify the agent behaves the way you want before you trust it with real work. Edit or add more in the editor.