Your candidate said 5 years of management. Their reference said individual contributor. Which screening vendor caught both?

Le problème des trois fournisseurs : personne ne relie les points

Voici comment se déroule la présélection des candidats dans la plupart des entreprises de marché intermédiaire aujourd’hui. Un recruteur utilise un outil pour la présélection ou les entretiens téléphoniques initiaux. Un autre outil gère les vérifications de références. Un troisième fournisseur effectue la vérification des antécédents. Chaque outil produit un rapport dans son propre format, derrière sa propre connexion, sur son propre cycle de facturation.

Now imagine a candidate claims six years of experience as a Senior Marketing Manager at a well-known company. Their pre-screen goes well — they're articulate and confident. But a reference describes them as a Marketing Manager (not Senior) who managed a team of 8 with a $2.5M budget — not the team of 15 and $5M budget the candidate claimed. Meanwhile, the background check shows they were actually at the company for four years, not six.

Three data points. Three different stories. Three separate vendor reports that nobody has time to cross-reference line by line. The recruiter glances at each report individually, sees nothing alarming in isolation, and moves the candidate forward. The discrepancies go undetected.

This isn't a hypothetical. Resume fraud affects a significant portion of candidates — Robert Half has reported that a majority of managers have caught candidates misrepresenting themselves on resumes. The problem isn't that screening tools don't work. The problem is that they don't talk to each other.

Comment fonctionne l’Evidence Intelligence

Virvell's Evidence Intelligence compares data across all three screening modules and generates an intelligence report that shows verified facts (where all sources agree), contradictions (where sources conflict, with specific details and severity ratings), and items warranting attention (patterns of misrepresentation across multiple categories).

Each flagged discrepancy includes the specific source that conflicts, what was claimed versus what was found, and a severity level — giving hiring managers concrete talking points for final interviews rather than vague concerns.

Pourquoi les présélections sont la fondation de tout

La présélection traditionnelle est un processus manuel et chronophage. Les recruteurs passent des heures en entretiens téléphoniques qui pourraient être automatisés — à poser sans cesse les mêmes questions de qualification, à tenter de composer avec les fuseaux horaires, à prendre des notes qui finissent dans des feuilles de calcul ou des champs de commentaires d’un SIRC que personne ne lit.

Les entretiens de présélection par IA de Virvell mènent des conversations vocales structurées avec les candidats — en les interrogeant sur leur expérience, leurs qualifications, leurs attentes salariales et les compétences propres au poste. L’IA génère une transcription et un résumé complets pour chaque candidat.

Mais la présélection fait quelque chose de plus important que de filtrer : elle capte ce que le candidat affirme à son propre sujet, dans ses propres mots. Années d’expérience. Titres occupés. Tailles d’équipes gérées. Responsabilité budgétaire. Compétences techniques. Motifs de départ. Ces affirmations précises deviennent les données structurées que les références et les vérifications d’antécédents confirment ou contredisent.

C’est pourquoi les présélections ne sont pas qu’une étape de tri — elles constituent la fondation de données qui alimente l’Evidence Intelligence. Sans un relevé structuré de ce que le candidat a affirmé, il n’y a rien à recouper. La présélection crée la référence de base; les vérifications de références et d’antécédents fournissent la couche de vérification.

Approfondissement : Comment fonctionnent les entretiens de présélection par IA de Virvell — des appels vocaux conformes à la juridiction qui captent les affirmations structurées des candidats en 5 à 15 minutes.

Ce que la détection des écarts repère réellement

Here's a real example of how Evidence Intelligence works across all three screening modules. The candidate applied for a Marketing Manager role and completed a pre-screen interview, had references contacted, and passed through background verification.

No single screening tool catches this pattern. A reference check alone might note the team size discrepancy but wouldn't know the candidate claimed something different in their pre-screen. A background check alone would catch the tenure gap but wouldn't know about the title inflation. Only a platform that processes all three data streams can connect these dots and reveal the pattern of systematic embellishment.

Importantly, Virvell does not score or rank Sarah. It doesn't recommend hire or don't hire. It presents the findings with specific sources and severity levels, and the hiring team makes the decision. Some discrepancies are dealbreakers. Others are understandable. That judgment belongs to humans.

Pourquoi les outils de vérification de références à fournisseur unique ne peuvent pas le faire

Le coût d’une présélection déconnectée

Beyond direct cost savings, the bundled platform eliminates hidden costs that don't show up in vendor invoices: recruiter time spent logging into three different systems, manually comparing reports from different formats, chasing down discrepancies that a connected platform would catch automatically, and managing three separate vendor relationships with three sets of contracts, renewals, and support contacts.

For a team doing 150 hires per year, the three-vendor approach costs $12,000-36,750 annually in vendor fees alone, plus hundreds of hours of manual comparison work. Virvell's Starter tier covers 15 credits/month at $699/month — with automatic discrepancy detection included.

Comment la plateforme devient plus intelligente à chaque présélection

Most screening tools run the same static scripts regardless of how many candidates they process. Virvell works differently. After every completed call, the AI analyzes the full transcript and generates 18 fields of structured intelligence: engagement level, sentiment scores, question effectiveness, coaching indicators, hesitation patterns, topic coverage, drop-off points, and more.

Before each new call, the system queries the last 20 completed conversations for the same role type, identifies patterns in what worked and what didn't, and injects a "Lessons Learned" section into the AI's instructions. The result: a screening conversation that adapts based on real data from real calls.

Pour les entretiens de présélection

The AI learns which questions surface genuinely useful signal for each role type and which ones generate generic answers that don't help distinguish candidates. It adapts follow-up probing based on patterns it has observed in how strong and weak candidates respond differently to the same topics. The screening conversation improves with every candidate who goes through it.

Pour les vérifications de références

Reference calls present a unique challenge: references often give rehearsed or scripted responses. The AI learns to detect these patterns and adapts its probing techniques accordingly. It identifies which follow-up questions consistently get references to move past surface-level endorsements and provide substantive, specific feedback. The more references the system conducts, the harder it becomes to game.

L’effet cumulatif

Because Virvell processes pre-screens, references, and background checks through one platform, the self-improving capability compounds across all three modules. Patterns detected in pre-screens inform how references are conducted. Discrepancies found in background checks refine what the AI probes for in future pre-screens. Single-service competitors can only improve within their one data type. Virvell improves across all three.

This is a structural advantage that grows over time. Every screening that runs through Virvell makes the platform better for every future screening. The AI collects data and identifies patterns for human review. It does not score candidates, make hiring recommendations, or auto-reject. The improvement is in conversation quality: better questions, better probing, better signal extraction.

Foire aux questions