Banking, Vietnam · Outbound voice, Talent Acquisition · Cloud · Pilot to October 2026
Voice AI for Candidate Prescreen and Interview Scheduling
Mass-hire screening at 4,000 to 5,000 candidates a year, without adding recruiters.
- 282
- candidates screened by the AI in the pilot
- 62%
- contact rate across the pilot
- 8/10
- quality, rated by VPBank’s own HR and data teams
The client
VPBank is one of Vietnam’s largest private banks. Its Talent Acquisition Centre hires about 4,000 to 5,000 candidates a year into mass roles, which means fresh graduate and junior positions. VPBank’s enterprise data and analytics division owns the AI platform.
The problem
For mass hiring, a recruiter had to call each candidate, ask the same 3 to 5 screening questions, judge the answers, and then book the ones who passed into an interview slot. At 4,000 to 5,000 candidates a year that consumed the Talent Acquisition team’s capacity. Scheduling was worse: recruiters booked interviews by hand over email, which left gaps between slots that no candidate filled, so interviewer time went to waste. VPBank wanted to raise recruiter productivity and modernize the candidate experience without adding recruiters. VPBank framed it as a change in how it hires, not as a software purchase.
What Revve built
- Prescreen agent: Calls the candidate, asks the scripted questions, handles off-script questions, and scores the answers against criteria and weights that HR and Revve designed together. Returns the transcript, the audio, and a scored evaluation with reasons. HR can change the criteria and re-score.
- Scheduling agent: Calls a batch of candidates and books them into consecutive interview slots, so it closes the gaps that the manual email flow left open.
- Integrations: SAP SuccessFactors (REC module). Pull candidate lists in bulk, push evaluations back.
- Guardrails: The agent stays on the approved script. It discloses that it is an AI. Candidates get a notice email before the call.
- Human handoff: Every candidate the agent passes still goes to a human Talent Acquisition interview. The agent screens; it does not decide.
Results
Measured from the Revve platform, 5 June to 31 August 2026. Test calls excluded.
| Metric | Value |
|---|---|
| Candidates called | 282 unique candidates, 1,276 calls |
| Contact rate across the pilot | 62% |
| Average length of a call that reached a candidate | 2.2 to 2.6 minutes |
| Roles covered | 3: mortgage applicants, call centre CSR, and SME relationship staff |
The pilot scaled every month
- June 2026: 68 candidates called, 407 calls, 72.7% contact rate
- July 2026: 72 candidates called, 310 calls, 59.0% contact rate
- August 2026: 153 candidates called, 559 calls, 55.8% contact rate
What VPBank’s own HR and data teams measured
- Conversation quality 80 to 90% human-like, rated about 8/10, which HR judged good enough to run in production (July 2026).
- In an early batch, all 8 candidates who completed the script also passed the human interview.
- 20+ candidates passed screening as of mid-August.
In their words
Revve is an enabler platform. It lets us roll out a new voice AI use case in days instead of months.
What is next
HR wants to deploy the agent across many more roles, calling from a VPBank-branded number. The team also wants to add the Revve email agent as a second channel for reaching candidates. Beyond HR, VPBank is discussing Revve as a group-level platform at Chief Data Officer level.
