The industry-specific problem
Recruiters spend time on mismatched applicants, but an aggressive automated technical quiz can screen out qualified people with disabilities, poor phone connectivity or unfamiliar accents. Mandatory job criteria must be separated from subjective communication impressions.
Step-by-step workflow
- Read the current Bullhorn job requirements in n8n and version a recruiter-approved scoring rubric.
- Invite the applicant to a Vapi interview with appropriate disclosure, consent and a text or human alternative.
- Ask consistent job-relevant questions; avoid protected traits and unsupported resume-verification claims.
- Use a supported Claude model to map quoted responses to rubric evidence. Replace the retired Claude 3.5 Sonnet dependency and evaluate bias and transcription error.
- Write a Bullhorn scorecard with missing evidence and accommodation requests. A recruiter makes the progression decision.
Visual workflow
Bullhorn job rubric → n8n interview invitation → Vapi consent/alternative
→ transcript evidence → Claude rubric mapping → recruiter Bullhorn review
Specific configuration example
This n8n Code-node fragment consumes a normalized, authenticated event from preceding integration steps. It demonstrates the decision boundary, not a complete API client. Fields are workflow-local mappings, not claimed vendor field names. Upstream credentials, schema validation, permissions, durable reservations and provider operations must be implemented separately. Numeric policy examples are illustrative, not Imperio service terms.
// n8n Code node: evidence score is not a rejection decision
const interview = $json;
return [{ json: { candidateId: interview.bullhorn_candidate_id,
rubricVersion: interview.rubric_version,
technicalEvidence: interview.answers.map(a => ({
criterion: a.criterion_id, quote: a.verbatim_answer,
needsClarification: a.transcription_uncertain })),
decision: "recruiter_review", scoreAccent: false,
accommodationRequested: interview.accommodation_requested === true } }];
Outcome to measure
Measure recruiter review minutes, qualified applicants advanced and disagreement with human assessment. Illustrative: five minutes removed from 100 initial screens frees about eight hours, subject to accessibility and selection-quality checks.
Compared with manual work or other tools
Application knockout questions are cheaper for factual prerequisites. Voice adds conversational clarification but cannot establish technical competence as reliably as validated work samples and human review.
Work with Imperio Automation
Ask Imperio to pilot an accessible Bullhorn screening assistant with evidence-based recruiter scorecards.
Frequently asked questions
Can it reject applicants below a model score?
Keep decisions under human review, with applicable employment-law and bias assessments.
Can it verify a claimed certification verbally?
A statement is not verification; use an authorized verification process.
What about speech disabilities or bad audio?
Offer an alternative and flag missing evidence without treating it as lack of skill.