For anyone running a live voice agent

Your dashboard says thecall was resolved. Was it?

Voice agents log calls as successful that quoted a wrong balance, never handed off an angry caller, or ended politely with the goal unmet. This auditor grades a batch of your call transcripts, catches those calls, and where any call is short names the one to review first — with a trust gate that forces MISHANDLED on any fabricated fact or failed handoff, however polished the rest of the call.

Get the Auditor — $129one-time · instant download · yours to keep
Six deliverables · runnable
Call Audit workbook (.xlsx)
excel
Runnable Python engine
cli
Four Claude Skills
skills
Exhaustive verifier (62,208-combo)
proof
Call QA Marking Playbook
docx
Fix & Re-Audit Runbook
docx
Works alongside
Voice Agent Deployment Kit · AI Agent Go-Live Readiness Gate
01.The Problem

The dashboard tracks whether the call ended — not whether it went well.

Confident

and wrong. A voice agent states a balance, a policy, or a price it could not know, in a natural voice a caller can't fact-check in real time. It's the failure mode QA teams rank most dangerous — and it logs as a normal call.

No handoff

A frustrated caller needed a human and the agent looped, faked competence, or dead-ended. Every missed escalation is a task the AI couldn't finish — and a customer who remembers it.

Polite,

unresolved. The call ended courteously with the caller's goal unmet. Containment counts it a win; the caller calls back. Transcript-only dashboards miss it because the words sound fine.

02.See It Work

Nine signals per call, a craft MIN, and one trust gate that outranks a clean score.

Live demo — the shipped 6-call sample. Tap a call, change its marks, watch the trust gate work.
CALL-1042
Taskw22
Groundingw20 · trust
Escalationw20 · trust
Consent to callw4 · policy
Recording consentw4 · policy
Identity disclosuresw4 · policy
AI disclosurew4 · policy
Handlingw12
Tonew10

Applicable policy and disclosure steps followed is four separate controls (the tiles tagged policy): consent to place or automate the call, recording consent, caller-identity or telemarketing disclosures, and AI-specific disclosure. Applicability depends on call type, direction, jurisdiction, purpose, technology and industry. Each records what the transcript and supplied context show: 0 = no evidence of the step identified in the transcript or supplied context; 1 = partial or incomplete evidence of the step; 2 = evidence of the step identified and recorded for review; NA = Not assessed. A 0 on any one of the four sets MISHANDLED on its own, as a 0 on any craft signal does. The top mark records the user’s assessment; it is not a certification of compliance, validity or legal sufficiency.

Weakest craft mark 2/2Score: 80Craft-derived band: RESOLVEDFinal result: MISHANDLEDPriority control: Factual grounding
Trust gate: Fired
Triggering marks: Factual grounding = 0
Graduated effect: A trust mark of 0 forces MISHANDLED
Final result: MISHANDLED

Your weakest craft mark falls in the RESOLVED AS DESCRIBED band, but Factual grounding is marked “a fabricated fact stated as true (a made-up balance, policy, price, or feature),” which triggers this instrument’s graduated Trust gate and forces the final result to MISHANDLED. The gate can worsen—but never improve—the craft-derived band.The result name is RESOLVED. A result sentence writes that band as RESOLVED AS DESCRIBED, because the sentence reports the call as the reviewer marked it.

The policy-and-disclosure mark reflects the reviewer’s assessment of the transcript and supplied context. It is not a determination that the call complied with the TCPA, recording-consent laws, telemarketing rules or other applicable requirements.

Batch verdict
MISHANDLING FOUND
Mishandle rate
67% (4 of 6)
Review first
CALL-1051 — Escalation correctness

The policy-and-disclosure mark reflects the reviewer’s assessment of the transcript and supplied context. It is not a determination that the call complied with the TCPA, recording-consent laws, telemarketing rules or other applicable requirements.

Same signals, weights, and graduated Trust gate as the workbook and the runnable engine — one scoring model, three surfaces. Deterministic and offline: it takes marks only, never transcript text, grades calls that already happened, connects to no live agent, and scores no people. Refresh resets the demo; the kit is yours to keep.

Diagram of the Voice-Agent Conversation QA & Escalation Auditor: nine craft-and-trust signals and a MISHANDLED verdict forced by factual grounding at 0 despite seven clean craft signals.
Shareable diagram

How the gate works, in one image

How the Voice-Agent Conversation QA Auditor grades a call's craft and trust separately and lets a bad trust signal override clean craft — the same math the demo runs, as a diagram you can share or embed anywhere.

View & embed the full diagram
03.The Runnable Engine

One command grades the whole batch — this output is from the shipped sample, run and verified.

python3 voice_qa_auditor.py sample_calls.csv
VOICE-AGENT CONVERSATION QA & ESCALATION AUDIT (v1.2)
==========================================================================
Signals (weights):
  Task resolution 22
  Factual grounding 20
  Escalation correctness 20
  Applicable policy and disclosure steps followed: consent to place or
  automate the call 4
  Applicable policy and disclosure steps followed: recording consent 4
  Applicable policy and disclosure steps followed: caller-identity or
  telemarketing disclosures 4
  Applicable policy and disclosure steps followed: AI-specific disclosure
  4
  Conversation handling 12
  Tone appropriateness 10
Result = the craft-derived band (the weakest assessed mark across task
resolution, the four applicable policy and disclosure controls,
conversation handling and tone), then the graduated Trust gate on factual
grounding and escalation correctness: a trust mark of 1 caps the final
result at SOFT-HANDLED, a trust mark of 0 forces MISHANDLED. A 0 on any
one of the four policy and disclosure controls sets MISHANDLED on its own.
A policy and disclosure control marked NA (Not assessed) is left out of
the score and the craft minimum and prevents RESOLVED. The result name is
RESOLVED. A result sentence writes that band as RESOLVED AS DESCRIBED,
because the sentence reports the call as the reviewer marked it.
--------------------------------------------------------------------------
CALL-1042     80  MISHANDLED
    Balance inquiry - agent stated a specific balance it could not v
    Trust gate: Fired
      Triggering marks: Factual grounding = 0
      Graduated effect: A trust mark of 0 forces MISHANDLED
      Final result: MISHANDLED
    Your weakest craft mark falls in the RESOLVED AS DESCRIBED band, but
    Factual grounding is marked “a fabricated fact stated as true (a
    made-up balance, policy, price, or feature),” which triggers this
    instrument’s graduated Trust gate and forces the final result to
    MISHANDLED. The gate can worsen—but never improve—the craft-derived
    band.
    Priority control: Factual grounding
    Note: The policy-and-disclosure mark reflects the reviewer’s
    assessment of the transcript and supplied context. It is not a
    determination that the call complied with the TCPA, recording-consent
    laws, telemarketing rules or other applicable requirements.
CALL-1043    100  RESOLVED
    Appointment reschedule - smooth, confirmed new time
    Trust gate: Not fired
    No scored control is below the top mark.
    Note: The policy-and-disclosure mark reflects the reviewer’s
    assessment of the transcript and supplied context. It is not a
    determination that the call complied with the TCPA, recording-consent
    laws, telemarketing rules or other applicable requirements.
CALL-1051     39  MISHANDLED
    Billing dispute - agent looped, never handed off the angry calle
    Trust gate: Fired — final result already at the same or a lower band; no further change.
      Triggering marks: Escalation correctness = 0
      Graduated effect: A trust mark of 0 forces MISHANDLED
      Final result: MISHANDLED
    Priority control: Escalation correctness
    Note: The policy-and-disclosure mark reflects the reviewer’s
    assessment of the transcript and supplied context. It is not a
    determination that the call complied with the TCPA, recording-consent
    laws, telemarketing rules or other applicable requirements.
CALL-1067     91  MISHANDLED
    Order status - resolved; no recording notice appears in the tran
    Trust gate: Not fired
    Priority control: Applicable policy and disclosure steps followed: recording consent
    Note: The policy-and-disclosure mark reflects the reviewer’s
    assessment of the transcript and supplied context. It is not a
    determination that the call complied with the TCPA, recording-consent
    laws, telemarketing rules or other applicable requirements.
CALL-1078     62  MISHANDLED
    Return request - polite throughout but caller hung up unresolved
    Trust gate: Fired — final result already at the same or a lower band; no further change.
      Triggering marks: Escalation correctness = 1
      Graduated effect: A trust mark of 1 caps the final result at SOFT-HANDLED
      Final result: MISHANDLED
    Priority control: Task resolution
    Note: The policy-and-disclosure mark reflects the reviewer’s
    assessment of the transcript and supplied context. It is not a
    determination that the call complied with the TCPA, recording-consent
    laws, telemarketing rules or other applicable requirements.
CALL-1090     90  SOFT-HANDLED
    Plan change - flawless handling but quoted an unverified promo r
    Trust gate: Fired
      Triggering marks: Factual grounding = 1
      Graduated effect: A trust mark of 1 caps the final result at SOFT-HANDLED
      Final result: SOFT-HANDLED
    Your weakest craft mark falls in the RESOLVED AS DESCRIBED band, but
    Factual grounding is marked “an unverifiable but low-stakes claim,”
    which triggers this instrument’s graduated Trust gate and caps the
    final result at SOFT-HANDLED. The gate can worsen—but never
    improve—the craft-derived band.
    Priority control: Factual grounding
    Note: The policy-and-disclosure mark reflects the reviewer’s
    assessment of the transcript and supplied context. It is not a
    determination that the call complied with the TCPA, recording-consent
    laws, telemarketing rules or other applicable requirements.
--------------------------------------------------------------------------
BATCH VERDICT: MISHANDLING FOUND
Mishandle rate: 67% (4 of 6 calls MISHANDLED; 1 RESOLVED)
Review first: CALL-1051 -> Escalation correctness
Note: The policy-and-disclosure mark reflects the reviewer’s assessment of
the transcript and supplied context. It is not a determination that the
call complied with the TCPA, recording-consent laws, telemarketing rules
or other applicable requirements.
==========================================================================
Deterministic and offline. Grades marks a human recorded from transcripts
of calls that already happened; takes marks only, connects to no live
agent, sends nothing, scores no people. Not legal, safety, or compliance
advice.

Read CALL-1042: it scores 80 — and reads MISHANDLED anyway, because the trust gate caught a balance the agent couldn't verify. And CALL-1090 has flawless craft yet reads SOFT-HANDLED, because it quoted an unverified rate — the gate doing work a quality average would hide. Zero dependencies, Python 3.8+, fully offline; the workbook carries the same scoring as formulas, and a 62,208-combination verifier ships in the zip that shows the trust gate worsens the craft-derived band on 3,288 of those combinations.

04.The Standard

Two layers, each doing distinct work, and a verdict from your own confirmed marks.

Craft MIN

The weakest assessed mark across seven craft signals — task resolution, the four applicable policy and disclosure controls, conversation handling, tone — sets the craft-derived band, because one broken part of the interaction sinks the call. Each policy and disclosure control also takes Not assessed, which is left out of the score and prevents RESOLVED. A 0 on any one of the four sets MISHANDLED on its own.

The trust gate

The two trust signals — factual grounding and escalation correctness — then worsen that base. A 1 caps a call to SOFT-HANDLED even at perfect craft; a 0 forces MISHANDLED. One graduated gate, worsen-only, and it never lifts a verdict.

Your marks, your verdict

The four Claude Skills propose marks and cite the transcript turns; a human confirms every mark before the verdict stands. No hidden benchmark, no invented rate — the verdict comes from what the calls actually show.

05.What This Is — And Isn't

The post-launch QA layer, not a live monitor and not another agent.

What it is
  • A deterministic, offline QA read of a batch of call transcripts — per-call verdict, mishandle rate, and the one call to review first where any call is short.
  • Four Claude Skills that read your transcripts, mark each call with cited evidence, draft the fix, and assemble the report — for you to confirm and act on.
  • One scoring model on three surfaces — workbook, Python engine, and this page's demo — with a shipped verifier that sweeps every mark combination and shows the trust gate earns its place.
  • The measurement layer beside the AI Voice Agent Deployment Kit (build) and the AI Agent Go-Live Readiness Gate (pre-launch).
What it isn't
  • Not a live monitor or a call simulator — it grades calls that already happened, connects to no live agent, and dials nothing.
  • Not an auto-fixer. It drafts recommended changes — a prompt edit, a KB gap, an escalation rule — for a human to review and apply to their own agent.
  • Not a scorer of people. It grades the AI agent's conversation, never the caller and never the human who takes a handoff.
  • Not legal, safety, or compliance advice, and no resolution or containment outcome is guaranteed.
06.Who It's For

Anyone accountable for what a live voice agent says to customers.

Ops and CX leaders running a Vapi, Retell, ElevenLabs, or GoHighLevel voice agent in production
Agencies deploying voice agents for clients who need a defensible QA read, not a vendor dashboard
Founders who turned on a voice agent and want to know what it's actually saying on calls
Support teams whose containment rate looks great but callbacks and escalations are rising
Anyone who just shipped a voice agent and needs the first honest audit of a real call batch
Teams in regulated-adjacent contexts where a hallucinated fact on a call is a real liability
08.Common Questions

Direct answers on scope, the trust gate, and the calls this catches.

No. It is deterministic and fully offline. You supply transcripts of calls that already happened; you (or the four Claude Skills) mark nine signals per call (0, 1 or 2, with Not assessed offered on the four policy and disclosure controls), and the engine returns the verdict. It takes marks only, never transcript text. It connects to no live agent, dials nothing, and sends nothing.

Find the calls your
dashboard is hiding.

One audit tells you which calls your agent actually mishandled — and, where any call is short, the one to review first. One purchase, lifetime access, 12 months of updates. $129, once.

Sold by RedHub AI LLC · Secured by Stripe · redhub.ai