Your dashboard says the call went fine.This catches the phantom booking it missed.
Run a whole week of voice-agent calls through one deterministic pipeline. It scores every call from your own marks, forces the polished call built on a false confirmation to ESCALATE NOW, rolls up your containment and booking, names the one call to review first where any call is short, and exports the flags straight into GoHighLevel.
The worst voice-agent call looks great on the dashboard.
Grade a call. Watch the keystone gate override the score.
Each of the four call-record controls is marked on its own evidence. A mark on one says nothing about the other three. A control marked NA is left out of the score and is never counted as a failed control. While any of the four is NA, a call cannot read CALL RECORD CONTROLS PRESENT AS DESCRIBED; it reads NEEDS EYES. The top result also requires that none of the four call-record controls is marked 0; with one at 0 the result is NEEDS EYES. A top mark records the user’s assessment; it does not certify compliance with TCPA, recording-consent, AI-disclosure, or sector-specific law.
Your score falls in the CALL RECORD CONTROLS PRESENT AS DESCRIBED band, but False confirmation is marked On, which triggers this instrument’s Keystone gate and forces the final result to ESCALATE NOW. The gate can worsen—but never improve—the score-derived band.
This result is based only on user-supplied marks. RedHub did not review the recording, transcript, consent evidence, disclosure, or applicable law. The result does not determine whether the call was lawful or whether any notice or consent requirement was satisfied.
Toggle “False confirmation” on this 95/100 call: a call your dashboard loved is forced to ESCALATE NOW, because it told the caller something before verifying it. The score never earns its way out of the gate.
How the gate works, in one image
How the Post-Call Intelligence Pipeline turns 9 signals (5 call-quality signals and 4 call-record controls) and one keystone gate into one verdict — the same math the demo runs, as a diagram you can share or embed anywhere.
View & embed the full diagramOne command grades the week and writes the GHL export.
The same verdicts you see in the demo come from a zero-dependency Python pipeline you run on your own call log. Here is its actual output on the shipped eight-call sample:
POST-CALL INTELLIGENCE PIPELINE - WEEKLY RUN (v1.1.0)
============================================================================
PER-CALL VERDICTS
----------------------------------------------------------------------------
CALL-3001 100/100 CALL RECORD CONTROLS PRESENT AS DESCRIBED
Keystone gate: Clear
CALL-3002 95/100 ESCALATE NOW
Keystone gate: Fired—changed result
Trigger checks: False confirmation detected: yes; Required
escalation did not fire: no
Your score falls in the CALL RECORD CONTROLS PRESENT AS
DESCRIBED band, but False confirmation is marked On, which
triggers this instrument’s Keystone gate and forces the final
result to ESCALATE NOW. The gate can worsen—but never
improve—the score-derived band.
CALL-3003 95/100 CALL RECORD CONTROLS PRESENT AS DESCRIBED
Keystone gate: Clear
CALL-3004 85/100 ESCALATE NOW
Keystone gate: Fired—changed result
Trigger checks: False confirmation detected: no; Required
escalation did not fire: yes
Your score falls in the CALL RECORD CONTROLS PRESENT AS
DESCRIBED band, but Escalation required is marked On and
Escalation fired is marked Off, which triggers this instrument’s
Keystone gate and forces the final result to ESCALATE NOW. The
gate can worsen—but never improve—the score-derived band.
CALL-3005 73/100 NEEDS EYES
Keystone gate: Clear
CALL-3006 85/100 NEEDS EYES
Keystone gate: Clear
Your score falls in the CALL RECORD CONTROLS PRESENT AS
DESCRIBED band. However, AI-generated/artificial-voice
disclosure, where applicable is marked Not assessed. Under this
instrument’s stated rule, an applicability question that is not
assessed prevents the top result. Final result: NEEDS EYES.
Not assessed: AI-generated/artificial-voice disclosure, where
applicable
Applicability not assessed — no legal conclusion drawn.
CALL-3007 100/100 CALL RECORD CONTROLS PRESENT AS DESCRIBED
Keystone gate: Clear
CALL-3008 45/100 ESCALATE NOW
Keystone gate: Clear
WEEKLY ROLLUP
----------------------------------------------------------------------------
Calls graded ............ 8
ESCALATE NOW ............ 3 (escalate share 37.5%)
Contained end-to-end .... 5 / 8
Booked .................. 4 / 8
Dead-air flags .......... 2
Interruption flags ...... 2
False confirmations ..... 1
Failed handoffs ......... 1
Top coaching theme ...... verify before confirming
Review first ............ CALL-3008
BATCH VERDICT: CONTROL GAPS OPEN
RESULT NOTES
----------------------------------------------------------------------------
On every result above, per call and batch:
This result is based only on user-supplied marks. RedHub did not review
the recording, transcript, consent evidence, disclosure, or applicable
law. The result does not determine whether the call was lawful or whether
any notice or consent requirement was satisfied.
On every CALL RECORD CONTROLS PRESENT AS DESCRIBED result:
This result reflects the records and marks supplied by the user. It does
not determine that the call, recording, disclosure or outreach complied
with the TCPA, state recording laws or other applicable requirements.
How the top result is reached:
A score of 75 or more with the keystone gate clear and no call-record
control Not assessed. The top result also requires that none of the four
call-record controls is marked 0; with one at 0 the result is NEEDS EYES.
Not legal advice. Voice-agent calling and recording touch TCPA, state
AI-disclosure, and recording-consent rules that vary by state and vertical.
This pipeline grades completed calls from your own marks as an operational
artifact — it scores no person, connects to nothing, and confirms no
compliance status. Confirm your obligations with counsel.Three rules the pipeline enforces on every call.
The dashboard grades what it can measure
Your platform marks a call ‘resolved’ when it ends cleanly. It cannot see that the agent confirmed a booking the tool never made. This grades resolution the way a human would — did the caller get what they came for, without being told anything untrue.
One keystone can’t be averaged away
A single false confirmation or a dropped handoff forces ESCALATE NOW regardless of score, because the harm is local and a high average hides it. A 95/100 call still escalates — the number never earns its way out of the gate.
A pipeline, not a spot-check
It runs on your whole week: per-call verdicts, escalate-share rollup, containment and booking counts, the one call to review first where any call is short, and a field export that lands the flags back on the contact record in GoHighLevel.
A grading pipeline. Not a live monitor, not a certifier.
- A deterministic pipeline that scores completed calls from your own marks.
- A keystone gate that catches false confirmations and failed handoffs.
- A weekly rollup and a GoHighLevel-importable field export.
- Fully editable — weights, bands, and thresholds are yours to tune.
- A live monitor — it connects to no agent and reads no live call.
- A transcript reader on its own — you supply the marks.
- A compliance certification or a legal opinion.
- A tool that scores, ranks, or rates any person.
Not legal advice. Voice-agent calling and recording touch TCPA, state AI-disclosure, and recording-consent rules that vary by state and vertical. This pipeline grades completed calls from your own marks as an operational artifact — it scores no person, connects to nothing, and confirms no compliance status. Confirm your obligations with counsel.
Operators who put an AI on the phone and now own the calls it makes.
The measure step in a full voice-agent loop.
The craft-level spot-check auditor. This pipeline runs the weekly loop around it.
ViewBuild the call flows and compliance layer. This grades what they produce.
ViewThe chat/SMS sibling to your voice line — capture, qualify, book, hand off.
ViewThe parts worth pinning down before you buy.
The call that ends politely
with a promise you can’t keep.
Find it every Friday, before the caller shows up to an appointment that isn’t there. One purchase, lifetime access, 12 months of updates. $129, once.
Not legal advice. Voice-agent calling and recording touch TCPA, state AI-disclosure, and recording-consent rules that vary by state and vertical. This pipeline grades completed calls from your own marks as an operational artifact — it scores no person, connects to nothing, and confirms no compliance status. Confirm your obligations with counsel.
Sold by RedHub AI LLC · Secured by Stripe · redhub.ai