Async screening
Collecting structured answers from candidates on their own time (forms, voice prompts, short video tasks, or chat-style bots) before a live recruiter call, so high-volume funnels stay fair and schedulable.
Michal Juhas · Last reviewed May 2, 2026
Who this is for
TA teams handling bursty inbound after campaigns or high-volume reqs who still want human judgment on finalists.
In practice
- Time-box tasks: five to ten focused minutes beats thirty-minute homework unless comped and justified.
- Publish rubrics: tell candidates what strong looks like without leaking answers.
- Route, do not auto-reject: let models tag "needs human" for borderline rows.
Where it breaks
Over-trusting bot scores, neglecting mobile UX, or skipping calibration across recruiters so the async step encodes bias from day one.
Async versus live phone screen
| Mode | Strength | Risk |
|---|---|---|
| Async | Schedules breathe, more coverage | Drop-off if friction is high |
| Live | Nuance and rapport | Harder to scale |
Related on this site
- Glossary: Scorecard, Structured output, Workflow automation
- Blog: AI candidate screening
- Guides: Talent acquisition managers
- Course: Starting with AI: the foundations in recruiting
Frequently asked questions
Is async screening the same as an AI phone screen?
Sometimes. The async part is any self-serve step; AI may score short text or structured answers, or route you based on availability. The principle is timing, not the specific medium.
When is it worth adding?
When reply volume from workflow automation or campaigns would swamp calendars. Pair with a clear SLA so candidates know what happens next.
What should we avoid?
Opaque rejections, long unpaid assignments, and tasks unrelated to the role. Candidates should understand how data is used and who reviews it. Automated decisions without oversight risk fairness complaints and bad hires.
How does this relate to scorecards?
Feed the same scorecard traits into the async form so downstream interviewers see consistent signals. Use structured output patterns when models summarize answers.
What limits should TA communicate?
Accessibility (captions, keyboard-friendly forms), language options, and opt-out paths for sensitive questions. Document retention for recordings or transcripts.
Where can we learn more?
Read AI candidate screening, refine prompts with few-shot prompting, and practice funnel design in a workshop.