PROBLEM
Too much recruiter time went into admin, not hiring decisions
The team had product-market fit, but operations did not scale. Every vacancy generated dozens of CVs, and recruiters manually turned raw profiles into hiring-manager-ready notes. Candidate records were also enriched by hand, creating delays, inconsistency, and a 4-day lag before shortlists reached clients.
✕Recruiters spent 340+ hours/year on repetitive screening notes and CRM updates;
✕Time-to-shortlist averaged 4 days, slowing client response and offer speed;
✕Candidate data lived across ATS exports, email threads, and HubSpot fields with 18% missing attributes;
✕Hiring managers rated the process poorly: summaries varied by recruiter and lacked consistent decision criteria.
SOLUTION · 8 WEEKS
One workflow for candidate summaries, enrichment, and recruiter QA
We built an AI-assisted ops layer around the existing stack, not a separate product. The goal was simple: turn every incoming candidate profile into a structured summary, enrich the CRM automatically, and keep a human approval step only where it added value.
WEEKS 1–2
Process mapping and data model
Mapped the screening flow across ATS exports, email intake, and HubSpot; defined 24 required candidate fields and approval rules for recruiter review.
WEEKS 3–4
LLM screening summaries
Built prompt chains that converted CVs and application answers into 6-part recruiter summaries: fit, risks, seniority, skills, salary signals, and next-step recommendation.
WEEKS 5–6
CRM enrichment automation
Connected n8n to HubSpot and internal spreadsheets to auto-fill missing fields, normalize titles and locations, deduplicate records, and push alerts to Slack.
WEEKS 7–8
QA, guardrails, and reporting
Added confidence thresholds, human review for edge cases, prompt versioning, and dashboards for turnaround time, completion rate, and recruiter override frequency.
RESULT · 3 MONTHS
Recruiters moved faster, hiring managers got better shortlists
| METRIC | BEFORE | AFTER | CHANGE |
|---|---|---|---|
| Recruiter routine hours / year | 340 | 0 | −340h |
| Time-to-shortlist | 4 days | 1 day | −75% |
| Hiring manager NPS | 43 | 64 | +21 |
| CRM records auto-enriched | 0% | 92% | NEW FLOW |
| Recruiter override rate | 100% | 14% | −86% |
“Before Pifagor, our recruiters were acting like data-entry clerks with good instincts. In 8 weeks, we turned that into a 1-day shortlist process our hiring managers actually trust.”