How AI Resume Screening Breaks the Hiring Bottleneck

Описание и применение

Автоматизированный пайплайн на базе n8n и LLM заменил ручной triage резюме, сократив время рекрутеров, обеспечил объяснимые оценки и выявил предвзятость по работодателям. Пайплайн разделяет большие партии, проводит аудит и требует ручной проверки перед интервью.

Задача

Recruiting teams at mid‑size companies are overwhelmed by the volume of applications, and traditional ATS systems only filter by keyword, missing qualified candidates and increasing recruiter workload.

Результат

The pipeline cut recruiter triage time from a week to an afternoon, produced explainable scores that recruiters used to craft interview questions, and revealed bias toward prestigious employers, prompting prompt engineering and bias audits.

Рабочий процесс

• 1. Export applicant CSV from ATS
• 2. Ingest resumes (PDF/plain‑text) into n8n
• 3. Parse resumes into structured text
• 4. Apply LLM reasoning model against a skills matrix derived from the job description
• 5. Generate ranked shortlist with reasoning per candidate
• 6. Write results to Google Sheet for live recruiter view
• 7. Conduct bias audit by running parallel pipeline with employer names redacted
• 8. Review and adjust evaluation prompt quarterly
• 9. Include human review gate before advancing to interview

Доказательства

The reasoning model we used had a tendency to weight prestigious employer names when they appeared in a resume, even when the job description said nothing about company pedigree.

Источник

Batch processing 200 resumes in a single n8n execution hit memory limits we had not anticipated.

Источник

Recruiters who used the pipeline reported that the written reasoning attached to each candidate score was more useful than the score itself.

Источник