A reference check is a conversation with consequences. Automating the scheduling and collection can reduce friction, but the process still needs to respect the referee, the candidate and the employer who will interpret the result.
A reference workflow with consent and review
What SEEK says it changed
SEEK launched Voice AI Reference Checks across its Asia Pacific markets in 2025. Its Business Council case story says referees can complete checks through natural conversation, with reusable references, voice and text options, and identity verification. SEEK reports completion times fell by 50 percent.
That speed figure is reported by the company through the case story. The page does not describe the evaluation design or a comparison group, so treat it as a signal worth exploring rather than a forecast for another hiring process.
The sensitive part is what happens to the answer
A reference may contain subjective impressions, personal information and details that were never intended for a machine generated summary. Candidates and referees should understand how voice or text is captured, retained and shared. Employers also need to know what identity checks establish and what they do not.
An AI summary can make a conversation easier to scan, but it can also remove hesitation, nuance or disagreement. Hiring teams should be able to review the underlying evidence and avoid using an automated score as a substitute for judgement.
Questions before adopting voice AI
Map the whole path from invitation to decision. Make alternatives available for people who cannot or do not want to use voice. Decide how corrections are requested and how a candidate can challenge inaccurate information. Keep the human decision maker accountable.
- Explain recording, transcription, retention and access in plain language.
- Test accents, noise, pauses and non standard speech.
- Do not rank candidates from a generated narrative without validated criteria.
Want to explore how this could apply to your organisation?
Explore Private AI ↗