Resume Screening Workflow: Consistent Shortlists Without Reading 400 CVs
Resume screening automation applies the role's criteria — the same criteria, the same way — to every application, and explains each score so the recruiter can disagree. The workflow parses applications from the ATS, scores must-have and nice-to-have criteria with evidence quotes, flags claims to verify, and produces a ranked shortlist with the reasoning; the recruiter decides who to call.
Written by Max Zeshut
Founder at Agentmelt · Last updated Sep 11, 2026
The problem
A posting draws 300–800 applications. Recruiters skim for keywords, screening depends on who is doing it and how tired they are, strong candidates with unusual backgrounds are missed, and there is no record of why anyone was rejected.
What changes when it runs
Every application is scored against written criteria within minutes, with the evidence. The recruiter opens a ranked list with explanations, reviews the top band and a random sample of the rest, and every decision is logged with its reason. Time-to-shortlist falls from days to hours and screening is consistent across recruiters.
Trigger, then 8 steps
Trigger
ATS webhook (application received)
Greenhouse, Lever, Ashby, Workable or Teamtailor post each new application; a daily run catches any missed.
Receive the application
WebhookResume, cover letter, application answers and the job requisition id from the ATS.
Load the role criteria
Google SheetsMust-haves, nice-to-haves and disqualifiers for the role, written by the hiring manager in plain language and approved by HR; weights per criterion.
Parse the resume
Information ExtractorRoles, dates, employers, skills, education, certifications and locations into structured fields, with the source text kept for quoting.
Score against criteria with evidence
AI AgentEach criterion scored met / partially / not met with a quote from the application as evidence; the agent is instructed to ignore name, photo, age, gender, nationality and school prestige unless a certification is a legal requirement.
Flag verification items
CodeGaps, overlapping dates, claims that need checking (certifications, seniority) listed for the interview.
Build the ranked shortlist
CodeWeighted score and band (strong / possible / unlikely); disqualifier matches are listed separately with the rule, never silently dropped.
Present to the recruiter
HTTP RequestScores, bands and explanations written back to the ATS as a scorecard; the recruiter reviews the strong band plus a random sample from the others and moves candidates forward.
Audit and bias monitoring
Schedule TriggerWeekly: score distributions and pass-through rates by self-reported demographic group where legally collected; anomalies flagged to HR.
Data it touches
- ATS applications (Greenhouse, Lever, Ashby, Workable, Teamtailor)
- Role criteria sheet (hiring manager + HR)
- Company screening policy and legal constraints
Guardrails
- No candidate is rejected automatically; the workflow ranks and explains, recruiters decide.
- Protected characteristics are excluded from scoring by instruction and by stripping fields before the model sees them.
- Every score carries an evidence quote; unexplained scores are not shown.
- Pass-through rates by group are monitored where lawful, and criteria that skew are reviewed.
Consistency is the benefit; explanation is the safeguard
The value of automated screening is that criteria are applied the same way to application 1 and application 700. The risk is that a model applies criteria you did not write. Two design choices manage that: criteria are written in plain language by the hiring manager and HR, and every score comes with a quote from the application. A recruiter who sees “not met: no evidence of managing a team — resume lists individual contributor roles” can agree or overrule in seconds, and the overrule is logged.
Regulatory posture
Automated employment decision tools are regulated in several jurisdictions — bias audits and candidate notices in New York City, high-risk classification under the EU AI Act, and general anti-discrimination law everywhere. The workflow is designed as decision support: it does not reject anyone, it strips protected data before scoring, it logs every decision, and it monitors pass-through rates. Where a jurisdiction requires a formal audit or notice, that is configured in the custom build with your counsel.
Tools in the stack
| Tool | Role in this workflow |
|---|---|
| n8n | Intake, scoring pipeline, ATS write-back, monitoring |
| Claude | Parsing and evidence-based scoring |
| Greenhouse / Lever / Ashby | ATS of record |
| Google Sheets | Role criteria the hiring manager edits |
Want this running without building it?
Automation workflow
$247/month
We set up, host and maintain this workflow on n8n and connect it to your tools. Setup included, cancel monthly, you keep the JSON.
Custom build
$3,500–6,000 one-time
Your systems, your rules, your edge cases. A one-off build on Claude and n8n, delivered with documentation and a walkthrough.
Covers one ATS and up to 3,000 applications a month. Jurisdictions with specific automated-decision rules (NYC Local Law 144, EU AI Act high-risk provisions) need the custom build with the required audit and notice features.
Frequently asked questions
Does it reject candidates?
No. It ranks, explains and flags. Recruiters make every decision, and the decision and reason are recorded in the ATS.
How does it avoid bias?
By stripping names, photos and demographic signals before scoring, instructing the model to score only against the written criteria with evidence, monitoring pass-through rates by group where lawful, and keeping humans in every decision. It reduces inconsistency between recruiters, which is itself a bias source.
Which ATSs are supported?
Greenhouse, Lever, Ashby, Workable, Teamtailor and SmartRecruiters via API; others via email intake of applications.
Case study
AI HR Agent for Tech Company: 70% Faster Hiring Pipeline
How a 200-person tech company used an AI HR agent to screen resumes, schedule interviews, and cut time-to-hire from 45 days to 14 days.
Calculator
Put your own numbers in
Estimate the cost of the current process and what this workflow changes.
The pillar
AI HR Agent
Source candidates, screen applicants, and schedule interviews—no code required.