Key takeaways
- Use a written brief and definition of done.
- Keep approvals and escalation rules visible.
- Review quality before expanding the workflow.
Overview
Define observable criteria for context checks, accuracy, notes, escalation, and close. Sample ordinary, complex, and escalated calls and link every score to an observed moment.
Applying virtual assistant call quality sampling
Sample for risk, not convenience. Include routine calls, unusual requests, and escalations, then tie each score to a timestamp or observable phrase. A scorecard should distinguish an inaccurate answer from a good escalation made before the caller was promised something.
Role boundary and review
The assistant can assemble recordings or notes and mark evidence against the rubric. The service owner decides whether a pattern requires coaching, a script change, or a scope boundary. Keep disputed scores separate from confirmed findings so a small sample is not presented as a broad performance claim.
Use the [services library](/services) to define whether call support or admin support is in scope, and the [provider vetting guide](/provider-vetting) to supply continuity questions. [NIST guidance](https://www.nist.gov/itl/smallbusinesscyber) is a useful general safeguard reference.
Sampling record
Keep the sample rationale, rubric version, observed evidence, disputed scores, coaching action, and review date.

