Top candidates accept another offer within 10 days — slow screening loses talent. Without a structured workflow, CVs pile up in inboxes, screening is manual and inconsistent, interview scheduling is a 6-email thread, and offer letters are drafted from a Word template that the hiring manager forgets to send. The result: 40% of selected candidates drop out due to slow follow-up, and the cost-per-hire balloons.

The RGPD angle is critical too: candidate data scattered across inboxes and local drives violates the 30-day retention rule for non-recruited candidates and exposes the company to fines.

The job requisition is created from a structured form with exact fields: job_title (text), department (select), job_location (select: PAR/LYON/MRS/TLS/Remote), contract_type (select: CDI/CDD/Alternance/Freelance), headcount (number), salary_range_min (number), salary_range_max (number), required_skills (multiselect), nice_to_have_skills (multiselect), experience_years_min (number), posting_channels (multiselect: Career-site/LinkedIn/WelcomeC/Indeed), application_deadline (date), hiring_manager_id (lookup).

The workflow is structured as a decision tree with 3 branches (auto-reject, screening-top, screening-pending) after Step 3, and 8 steps end-to-end. Every action is appended to the timeline_recruitment audit subtable.

• Step 1 (Start — hiring manager): the hiring manager creates the requisition with job_title, required_skills and experience_years_min; the AI copilot drafts the job description from the role template and publishes it to posting_channels.

• Step 2 (Application intake): candidates apply via the portal; the workflow stores each application with CV, cover letter and structured fields.

• Step 3 (ai-agent — CV screening): the ai-agent module scores each CV against required_skills, nice_to_have_skills and experience_years_min, and ranks the applicants. Sub-branch A (auto-reject): if the score is below the threshold (e.g. 40%), send a polite rejection email and flag for RGPD 30-day purge. Sub-branch B (screening-top): if the score is above the shortlist threshold (e.g. 75%), auto-schedule interviews. Sub-branch C (screening-pending): if the score is in the middle band, route to the hiring manager for manual review.

• Step 4 (Interview scheduling): top candidates are auto-scheduled based on the hiring panel's calendars; the workflow sends invitations and reminders in 7 languages.

• Step 5 (Interview & scoring): each interviewer scores the candidate on a structured rubric; the ai-agent aggregates scores and flags divergences for the hiring panel.

• Step 6 (usePDF — offer): for the selected candidate, the workflow drafts the offer letter via usePDF with salary (within salary_range_min/max), start date and conditions; the hiring manager and HR validate it.

• Step 7 (Contract signature): the contract is sent for eIDAS-qualified electronic signature to the candidate and the legal representative; the workflow tracks signature status and sends reminders. Once signed, the workflow triggers the onboarding workflow.

• Step 8 (useArchive — RGPD 30-day): recruited candidates' data is archived in the employee record via useArchive; non-recruited candidates' data is auto-purged after 30 days to comply with RGPD.

Teams deploying Utilyx recruitment cut time-to-hire from 45 days to 18 days, reduce cost-per-hire by 35%, and eliminate RGPD non-compliance fines thanks to the automatic 30-day purge of non-recruited candidates via useArchive.

The ai-agent scores CVs and aggregates interview feedback — so the hiring panel focuses on decisions, not admin. The no-code designer lets HR adjust rubrics, scoring weights and approval chains without coding.