SEEK Pass
Roark x SEEK Pass: Scaling AI voice agent testing for reference checks across Asia-Pacific
SEEK Pass uses Roark to simulate and evaluate AI-powered Reference Check conversations in English, Thai and Indonesian.
- Nearly 4,000
- simulated calls completed since February 2026
- 3.7 to 4.7 out of 5
- coverage of required questions between March and October 2026
01 Overview
SEEK Pass, part of SEEK Group, a leading employment marketplace in Asia-Pacific.
Use case: Automated testing of AI-powered voice agents for reference checks.
Languages tested: English, Thai and Indonesian.
Roark is a voice and chat agent observability and simulation platform.
Roark products: simulation testing, custom evaluation metrics, scheduled regression runs.
02 Challenge
SEEK Pass uses AI voice agents to conduct reference checks, helping employers gather valuable candidate insights more efficiently.
Operating across multiple Asia-Pacific markets, means these agents need to deliver natural, consistent and locally appropriate conversations in different languages.
Previously, testing relied heavily on manual calls to evaluate agent performance across different scenarios and languages. This was time-consuming and difficult to scale, particularly when assessing conversational quality, voice consistency and coverage of required reference-check questions.
03 How they use Roark
SEEK Pass uses Roark to simulate and evaluate AI-powered Reference Check conversations in English, Thai and Indonesian.
The team has developed 48 simulated caller personas and 25 test scenarios to assess agent performance against custom metrics, including coverage of the required reference-check questions, toxicity detection and accent adherence.
Automated daily regression tests in English and Indonesian help monitor performance and identify potential issues, while more comprehensive testing supports quality assurance before agent updates are released.
This approach enables SEEK Pass to test conversations more consistently, identify opportunities for improvement and reduce reliance on manual testing.
04 Results
Nearly 4,000
simulated calls completed since February 2026, supporting ongoing testing and optimisation.
3.7 to 4.7 out of 5
Improved question coverage: Across SEEK's recurring reference-check test suites, coverage of required questions increased from 3.7 to 4.7 out of 5 between March and October 2026.
+370
automated scheduled test runs, with daily English and Indonesian regression testing continuing to monitor agent performance.
Together, these improvements have helped SEEK Pass reduce manual testing effort, improve quality assurance and build greater confidence of its AI-powered reference checks.
05 In their words
Roark has reduced the time we spent manually testing by a few hours per week, allowing us to focus on improving the product. With automated daily testing, we have greater confidence that our voice agents are working consistently and can identify issues earlier. The quality of the simulated conversations has also been impressive. Our Indonesian and Thai speaking colleagues commented on how natural and authentic the simulated users sound.
Bring a recording.
We’ll score it live.
See your own agent measured on the audio it actually produced, in the demo, in real time. Stop guessing whether your voice AI works.
Or get $50 in credit with a work email · read the docs · support@roark.ai