AI Quality Auditor

QA scorecards across every agent on your VICIdial box.

The AI reviews the call. You set the rubric. Every call either passes or it doesn’t.

Define weighted scorecards with scored, checklist, critical, and open-ended checkpoints. Set sampling rules - every sale call, ten random per agent per day, every fifth call on one campaign. The AI runs each recording against your rubric and returns a per-checkpoint score, a section rollup, a pass/fail verdict, and coaching notes - and the results persist so your QA trends hold up in a dispute or a compliance audit.

See Managed Floor See pricing

Included with Managed Floor Enterprise · not sold separately

from $2,900/mo with Managed Floor

4checkpoint types

4sampling strategies

6languages

How it works

One rubric. Every call that matches. The AI does the marking.

The Auditor layers a structured audit workflow on top of the transcription and analysis pipeline already on your box. You own the rubric; the AI does the scoring.

Versioned scorecards

Build sections and checkpoints once. Saving creates a new version; prior reviews stay tied to the rubric they were scored against - tighten a scorecard mid-quarter without rewriting history.

Four checkpoint types

Scored (0–10), checklist (yes/no/N-A), critical (pass/fail), and open-ended (AI free-text). One scorecard carries all four - disclosures as critical, pitch quality as scored, objection handling as open-ended.

Sampling rules, not every call

Filter by campaign, ingroup, list, agent, disposition, and length, then pick a strategy - all, N random per agent per day, a percentage, or every Nth. A daily budget cap keeps AI spend from compounding overnight.

Per-section rollups + coaching themes

Section scores are the primary lens - “Closing averaged 41% this week” tells a manager where to spend 20 minutes. Open-ended outputs roll up across the fleet so you see recurring gaps, not one-off failures.

Instant-kill compliance items

Mark any critical checkpoint as instant-kill: one failure fails the whole review regardless of other scores. Built for mandatory disclosures, TCPA opt-in language, and identity verification where partial credit is not an option.

Speaker diarization in every review

Agent and lead audio are split before the model scores. Talk-time split, interruption count, and longest monologue surface in every review - scores that depend on agent behavior are not confused by lead chatter.

Setup in three steps

Build a scorecard. Write a sampling rule. Let it run.

No integration work - the AI pipeline already runs on your server. You configure the rubric and rules in the dashboard; scoring starts on the next cadence tick.

Build your scorecard

Create sections (Compliance, Opening, Discovery, Closing), add checkpoints, set weights, the pass threshold, and which critical items are instant-kills. Saving versions it; prior reviews are untouched.

Write a sampling rule

Pick which calls match (campaign, disposition, agent, minimum length), choose a sampling strategy, set the cadence, and optionally cap the daily AI spend.

Read the reports

See recent failures, per-agent section trends, and coaching themes. Drill into any review for per-checkpoint scores, evidence quotes with seek-to-moment audio, and the full transcript.

Who it’s for

For floors where “did the agent say it right?” actually matters.

Compliance-heavy floors

Collections, healthcare, finance - prove the mini-Miranda, the disclosure, the consent language was read, on every call that matters.

QA managers scoring at scale

Replace spreadsheet spot-checks with consistent, sampled scoring across every agent - no more “we reviewed five calls last month.”

Coaching programs

Turn per-section scores and recurring themes into targeted 1:1s, instead of vague “be more confident” feedback.

A QA analyst listening to recorded agent calls at a workstation — lens clickk / Pexels

vicifast - qa · scorecard

scorecard  "Compliance + Pitch"   v4
  ▸ Mini-Miranda disclosure   CRITICAL · instant-kill
  ▸ Rebuttal attempted        SCORED 0–10
  ▸ Tone / professionalism    SCORED 0–10
sampling  every SALE  +  10 random / agent / day
 
review #8821   agent 1043
  Mini-Miranda ............ FAIL  → review failed
  section: Closing ........ 41%   ▼ coach

One rubric. Every matching call scored automatically. Instant-kill on the items you can’t miss.

Pricing

It comes with Managed Floor Enterprise.

The Auditor is not sold on its own at any price. Scorecards nobody is coached against change nothing, so it ships with the engineer who runs the calibration - priced on your hosting plan, across every server on your account. Scoring usage auto-bills from your wallet at the platform per-minute rate: no markup, no per-agent seat, no per-review charge.

From $2,900/mo, with Managed Floor Enterprise

That is the 25-agent band; it scales with your plan up to $27,900/mo at 350 agents. Scoring auto-bills your wallet at the platform per-minute rate (a plan’s AI Call Analysis minutes don’t apply), with no markup on minutes. See the full price table.

Unlimited scorecards and sampling rules on your server
Per-checkpoint AI scoring, section rollups, pass/fail verdict, and coaching observations
Full transcript + speaker diarization in every review (agent vs lead split, talk %, interrupts)
A QA-reviewer sub-user role - add your QA team without granting full server access
Versioned scorecards - a review from six months ago still references the exact rubric it used

FAQ

Quality Auditor questions

AI Call Analysis is the per-recording instrument - pick one recording, get a transcript, a summary, sentiment, a quality score, red flags and a suggested disposition. AI Quality Auditor is the QA program on top: you define a scorecard and a sampling rule, and the platform automatically selects and scores calls across your fleet, then tracks the trend. The instrument versus the program that runs it systematically.

No. AI Call Analysis is available without the Quality Auditor. The Auditor uses the same transcription pipeline and is paid from the same wallet, but the analysis minutes a plan includes don’t count toward it. The Auditor is not sold on its own - it comes with Managed Floor Enterprise, which enables the scorecard builder, sampling rules, and QA reporting, and puts a named engineer on the coaching that makes the scores worth having.

Yes. Each checkpoint carries a free-text criteria field telling the model exactly what counts as full, partial, or zero marks - the more specific you are, the more consistent the scores. You can re-run any review after editing the scorecard; the new run is a fresh row and the original is preserved.

Language is auto-detected per recording - English, Spanish, Portuguese, French, German, and Hindi are supported. The scoring prompt is language-specific, so the model reads the transcript in the language it was spoken and the scores stay accurate.

The recording stays on your server. The platform pulls it through an authenticated pipeline in the same region, transcribes and scores it, and writes the review rows back to your database - no third-party storage of your audio. On export, optional PII redaction blurs SSNs, card numbers, phones, and emails before any transcript text leaves the platform.

Looking for per-recording transcription, sentiment, and suggested dispositions without a scorecard? See AI Call Analysis →

Your QA rubric. Every call that matches. Zero manual pulls.

It comes with Managed Floor Enterprise, from $2,900/mo on a 25-agent plan - scoring, calibration, and the engineer who runs both.

See Managed Floor pricing Tell us about your floor

How you get it

AI Quality Auditor comes with Managed Floor. It isn’t sold separately.

Not as a paywall — because it doesn’t work unattended. Scorecards nobody reads are shelfware, and an agent who is never coached on them scores the same next month. The tool is only half of it; the other half is an engineer whose job is to act on what it finds. So we stopped selling the half that doesn’t work on its own.

Managed Floor Essentials

Not included at this tier

Managed Floor Enterprise

Included

See Managed Floor pricing Or tell us about your floor

Already a Managed Floor customer? It’s switched on in your portal — no extra charge, nothing to add.