Installed snapshot · independent audit

After-Hours Lead Capture: strong pattern, dangerous callback loop.

A node-level review of Votel’s v5.1 residential-services snapshot: caller lookup, live forwarding, structured capture, post-call AI, callback automation, and the repairs required before activation.

Votel library v5.1Reviewed Sep 18, 2026Agency-level installNo live channels used
VOTEL SNAPSHOT AUDIT07 / RESIDENTIAL SERVICES

After-Hours
Lead Capture AI

Installed agent1
Agent workflows3
Workflow nodes32
Connected edges35
Launch statusNeeds rebuild
3Installed workflows
32Real canvas nodes
7Open outputs
12Audit findings
0Active channels
01 · Executive verdict

Excellent building blocks. The callback trigger needs an emergency brake.

This is one of the clearest demonstrations of Votel’s native orchestration model: identify the caller, attempt a live handoff, capture details when the team cannot answer, analyze the completed call, and recover missed leads. But its Call Ended trigger watches inbound and outbound calls while the missed-lead tag remains attached. That means a callback can end, qualify as another ended call, and start the callback workflow again.

Reusable

Core pattern

Contact lookup, call forwarding outcomes, shared caller capture and post-call structured AI.

Stop

Callback cycle

Any-direction Call Ended plus a persistent missed-lead tag can repeatedly re-enter outbound callbacks.

Remap

Destinations

Forwarding numbers and internal-looking email recipients are not safe reusable defaults.

Do not activate this snapshot as installed. Replace destinations and fix the callback trigger, tag transition, and attempt limit first.
02 · Installed inventory

One agent contains three complementary flows.

AssetStatusVerified detail
After-Hours Lead Capture AIInactive

GPT-5.4 Mini, Katie by Cartesia, en-US, zero active channels, 320-word prompt.

AFTER HOURS16 nodes

Caller lookup, forwarding, hours flags, four-question capture, missed-lead tag and three notification outcomes.

AFTER CALL SUMMARY8 nodes

Any ended call enters missed-lead callback logic or GPT-4.1 Mini action extraction.

CALL BACK HANGUPS8 nodes

Callback greeting, second forwarding attempt, capture, notification and farewell.

Forwarding destinationsConfigured

Two forwarding nodes contain two active account-specific destinations each.

Email actions5 unresolved

Internal-looking messages target the contact and have no selected sender.

Knowledge / widgets / toolsNone

No knowledge base, widget, custom tools, or enabled built-in tools.

State distinction: the snapshot is installed and connected. Its destinations, senders, business variables, channel scope, callback controls, and customer/team message contracts are not launch-ready.
03 · Workflow map

What the three flows are trying to accomplish.

After-Hours Lead Capture AI
● INACTIVE

Workflows

AFTER HOURSAFTER CALL SUMMARYCALL BACK HANGUPS

Agent

Prompt BuilderTools & IntegrationsShared Knowledge
AFTER-HOURS LEAD RECOVERY · 3 FLOWS
Inbound conversationLookup or create caller
Forward the callAnswered, no answer, outside hours
AnsweredStatus email
MissedCapture + tag + notify
Call endedAny status · any direction
Missed lead?Callback or AI summary
Recover / report2 calls or action email
Sanitized workflow reconstruction. No forwarding destination or private account value is shown.

1 · AFTER HOURS

Conversation starts→Lookup/create contact→Live forwarding→Capture + tag + notify

The No Answer and Outside Hours outcomes share one capture sequence, then split into business-hours and after-hours notification copy.

2 · AFTER CALL SUMMARY

Any call ends→Missed lead?→2 outbound callbacksorAI action summary

The trigger includes the callback calls themselves. Without a status transition or idempotency guard, the recovery path can re-enter.

3 · CALL BACK HANGUPS

Callback answered→Reconnect live→Capture if unavailable→Notify + end

A useful separate conversation flow, but it inherits configured destinations and unresolved email delivery.

04 · Prompt & data contract

The “residential services” prompt is actually one waterproofing company.

The agent is named Sara and includes specific Central Iowa cities, patented-method claims, lifetime-warranty claims, an emergency-chat claim, and incomplete FAQ answers. Ten business variables are unresolved, and caller contact fields are used where business contact details should appear.

Good architecture

  • Lookup before capture reduces duplicate contacts.
  • First name, last name and phone persist to standard fields.
  • Structured post-call output separates actions from prose.
  • Forwarding exposes clear answered/no-answer/outside-hours outcomes.

Unsafe data contract

  • Ten business globals are unavailable.
  • Caller email and work phone are presented as company details.
  • No knowledge base supports the business claims.
  • The fourth capture question is not shown as a mapped contact field.
Library mismatch: the listing says an early hangup can trigger a callback text. The installed workflow uses outbound phone calls and contains no SMS action.
05 · Defect register

Twelve findings, prioritized by launch risk.

FindingSeverityWhy it matters
Callback workflow can re-trigger itselfCritical

Any-direction Call Ended plus a persistent missed-lead tag can create repeating calls and charges.

Configured forwarding destinationsCritical

A reusable snapshot should never ship live external phone destinations.

Team notifications target the callerHigh

Five internal-looking emails use the contact's email address.

No email senderHigh

Every promised email path remains undeliverable.

Promised callback text is absentHigh

The graph places outbound calls and contains no SMS node.

Call flow accepts chat and SMSHigh

Non-voice conversations can enter call-ID and forwarding nodes.

Waterproofing claims are hard-codedMedium

Other residential businesses inherit inaccurate claims and locations.

Caller data used as company dataMedium

The agent can repeat the caller's own phone/email as business contact information.

Ten globals are missingMedium

Business identity and service answers cannot resolve reliably.

No knowledge; incomplete FAQMedium

Several answers are blank or generic Yes/No statements.

Missed-lead state never clearsMedium

Contacts remain eligible for recovery logic indefinitely.

Identity and copy cleanupLow

Sara is hard-coded, variable spacing is broken, and no opening greeting is configured.

06 · Launch checklist

Make callbacks finite, destinations explicit, and messaging honest.

  • 1
    Scope the Call Ended trigger

    Missed inbound calls only, not every status and direction.

    Safety
  • 2
    Add attempt state and idempotency

    Transition the missed-lead status and enforce a hard callback cap.

    Safety
  • 3
    Clear forwarding destinations

    Require and verify every team number during setup.

    Safety
  • 4
    Match the callback promise

    Add SMS, change the listing, or clearly implement both channels.

    Required
  • 5
    Restrict the initial trigger to voice

    Or route non-voice conversations away from call-only nodes.

    Required
  • 6
    Separate internal and customer emails

    Give each message an intended recipient, sender, and privacy contract.

    Required
  • 7
    Replace hard-coded business claims

    Use verified variables and governed knowledge.

    Required
  • 8
    Correct the data model and copy

    Use company fields for company facts, verify all four questions, and configure greetings.

    Required
  • 9
    Test all paths synthetically

    Prove callback termination before attaching any live channel.

    Test

After-Hours Lead Capture audit complete.

The snapshot is installed, inventoried, traced, sanitized, scored, and added to the first-party comparison and defect matrices.

Return to the Snapshot Library →
Audit method: read-only agency-account review on September 18, 2026. No agent was activated, no simulator case was generated, and no email, call, SMS, or live contact was used.