> ## Documentation Index
> Fetch the complete documentation index at: https://docs.klariqo.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Human-agent QA

> How Klariqo reviews human-agent VICIdial calls from the recording against the same scorecard, and attaches QA results to the signed compliance record.

Human-agent QA reviews calls handled by your human agents on VICIdial. Klariqo reviews those calls from the call recordings.

The same scorecard QA and sentiment analysis apply. This lets you use one operating ruleset across AI calls and human-agent calls.

<Note>
  Setting this up? See [Connect VICIdial](/dialer-integrations/connect-vicidial). It needs only recording access, not the SIP voice setup.
</Note>

## How human-agent QA works

<Steps>
  <Step title="The human-agent call is recorded">
    The review starts from the customer's VICIdial call recording.
  </Step>

  <Step title="The recording is transcribed">
    Klariqo transcribes the recording and attributes turns to the agent or the customer.
  </Step>

  <Step title="The same scorecard is applied">
    The call is reviewed against your configured required elements and prohibited claims.
  </Step>

  <Step title="Sentiment is attached">
    Sentiment analysis is attached alongside the QA result.
  </Step>

  <Step title="The result joins the record">
    The QA result and sentiment become part of the signed compliance record.
  </Step>
</Steps>

## Role attribution caveat

Role attribution is inferred from the audio. It identifies which speaker is the agent and which speaker is the customer, but it is not perfect.

Attribution is less certain on a single-channel, mono recording than on a dual-channel recording. When a turn cannot be confidently attributed, the record flags it as `role_unattributed`.

<Warning>
  Do not treat role attribution as perfect. Review flagged turns when attribution confidence matters for your workflow.
</Warning>

## Same rules across AI and human calls

Human-agent QA uses the same scorecard structure as AI-call QA. Required elements, prohibited claims, critical checks, and enabled checks work the same way.

That gives you one review language for your operating rules, whether the call was handled by Klariqo AI or by a human agent on your dialer.

<CardGroup cols={2}>
  <Card title="Scorecards" icon="list-check" href="/call-qa/scorecards">
    See the operating rules human-agent calls are reviewed against.
  </Card>

  <Card title="QA results" icon="clipboard-check" href="/call-qa/qa-results">
    Learn how to read status, score, and compliance status.
  </Card>
</CardGroup>

<Warning>
  Human-agent QA supports evidence, provenance, and audit-readiness. It shows how a call matched your configured scorecard. It does not make you compliant, prove consent by itself, approve your scripts, or replace counsel review.
</Warning>