How a recruiter can screen 200 resumes in 10 minutes with an autonomous AI agent

The volume of applications received per job posting has exploded since online application platforms became mainstream. A posting on LinkedIn or Indeed for a mid-level management role now receives 150 to 300 applications within the first 48 hours according to a LinkedIn Talent Solutions 2025 study on HR automation. For an independent recruiter, manually screening these applications represents 12 to 20 hours of work per engagement.

Nadia, an independent recruiter in Casablanca specializing in tech and marketing profiles, deployed an autonomous AI agent built with n8n and Claude to automatically screen and score 200 applications in 10 minutes. Her productivity increased by 90%, freeing 15 hours per week reinvested into client relationships and sourcing rare profiles.

The Anatomy of Nadia's Screening Agent

The agent is built on n8n, an open-source automation platform that can be hosted on a European server (GDPR compliance). It automatically chains 7 steps without human intervention after launch.

Step 1: Application collection. The agent monitors a dedicated applications inbox via IMAP. Each email with a resume attachment automatically triggers the rest of the workflow.

Step 2: Resume content extraction. PDF or Word files are converted to structured text. Key information is extracted: education, experience, technical skills, languages, location, availability.

Step 3: Analysis and scoring by Claude. Extracted text is sent to Claude Sonnet 4.6 with a prompt containing the job description and Nadia's scoring criteria. Claude assigns a score from 0 to 100 on each criterion and generates a 3-line summary explaining the overall score.

Step 4: Discriminatory criteria filtering. An explicit filtering block removes references to age, gender, origin, and family status before sending the resume to Claude. Only professional skills and experience are transmitted for analysis.

Step 5: Tracking table population. Scoring results are automatically inserted into an Airtable table with candidate name, overall score, per-criterion scores, and the Claude summary.

Step 6: Shortlist ranking. Candidates above a defined score threshold (generally 70 out of 100) are automatically grouped in a "Shortlist" tab and sorted by descending score.

Step 7: Notification to Nadia. When screening is complete, Nadia receives a Slack message with the number of applications processed, the number of shortlisted profiles, and a link to the Airtable table.

Results After 3 Months

Nadia used this agent on 8 recruitment engagements since deployment, processing a total of 1,400 applications. Average screening time dropped from 16 hours to 1.5 hours per engagement (including shortlist review time), a 91% reduction.

Across the 8 engagements, she found high client satisfaction: hiring managers who received her shortlists reported that 82% of presented profiles met their expectations, versus 68% before the agent.

What Made the Project Possible: Ethics Scoping First

Nadia spent the first day of her IMPACT engagement defining the ethical rules for her agent, before starting any technical configuration. The list of excluded criteria, candidate information modalities, data hosting: these questions were addressed upfront to avoid legal or reputational risk.

This is the approach we recommend for any AI agent project involving personal data. The Diagnostic-Assessment phase of the IMPACT methodology systematically includes a GDPR and ethical risk analysis before deployment.

The TransformAudit at 1,490 euros covers this ethical and technical scoping and enables deploying a compliant, effective resume screening agent within 30 days.

If you are an independent recruiter, HR director, or recruitment manager and application volume is exceeding your processing capacity, contact us for a 30-minute call.