Lab / Decision dossier

Evidence snapshot · 2026-08-19 · No runtime tests

Who should answer the next contractor lead?

A quality-first decision report comparing packaged receptionists, reseller platforms, managed voice infrastructure, open source, and the current Twilio/OpenAI foundation.

54product rows screened
79repositories screened
12deep source audits
269qualifying opinion URLs
24later tests · not run

01 / Decision

The recommended operating path.

Capability scores and evidence confidence are separate. This decision is provisional until the later hands-on protocol is executed.

Primary path

Retell + a TJ-owned contractor control plane

It leads the fixed scorecard and has the strongest combined independent platform evidence without forcing TJ to rebuild the media pipeline. The recommendation buys the difficult realtime substrate while keeping contractor workflow and customer policy as owned IP.

Path: Use Retell for managed voice orchestration and telephony primitives; own tenant profiles, contractor policy, integrations, artifacts, evaluation, and fallback logic.

Sources: Retell · Retell · Retell · Retell · Retell · Retell

Fallback

Synthflow agency / white-label + external TJ billing

It is the fastest credible packaged-reseller path with meaningful workflow and agency controls. It loses to Retell on evidence-backed quality/control and carries agreement-specific cost and parent-account risk.

Main risk: Exact rates and parent/subaccount limits are agreement-specific; in-product Stripe reselling is scheduled for removal on 2026-09-15.

Sources: Synthflow · Synthflow · Synthflow · Synthflow · Synthflow · Synthflow

Best overall quality candidate

Retell + TJ control plane

Highest fixed score and strongest combined independent platform evidence; still provisional.

Best packaged reseller / white-label

Synthflow

Most complete public agency/subaccount/workflow story among evidence-rich candidates; external billing and agreement terms remain risks.

Best developer platform

Retell

Strong balance of managed realtime infrastructure, telephony, functions, webhooks, and independent experience.

Best open-source / self-hosted path

LiveKit Agents + Pipecat evaluation patterns

Best integrated runtime plus strongest composable framework, but not a complete receptionist and not the first production path.

Best hybrid using current foundation

Twilio ConversationRelay + OpenAI Realtime

Preserves TJ's Twilio/OpenAI knowledge and maximum workflow ownership, with a materially larger product/operations burden.

Best human-backed safety net

Smith.ai

Strongest packaged human escalation and staffed-operations benchmark for high-consequence calls.

Best FSM-native path

Use the customer's existing Housecall Pro, Jobber, ServiceTitan, or Workiz AI

Native customer and schedule context can dominate when the customer is already committed to one field-service suite; it is not a universal agency default.

02 / Current system

Useful foundation. Not a finished receptionist.

Proof level: source-verified narrow controller; no completed-call proof. Source presence is not completed-call behavior.

Retain and reuse

  • Confirm-gated invocation
  • Signed Twilio/OpenAI ingress
  • Opaque contextual briefs
  • Direct SIP TwiML
  • Realtime sideband control
  • Limited screener/hold state
  • Transcript turns
  • Recording callbacks
  • Summaries
  • Seven-day artifact retention

Missing product controls

  • Structured contractor lead actions
  • Calendar/CRM/SMS/email tools
  • Transfer/REFER
  • Voicemail
  • Bounded hold timeout
  • SIP-failure fallback
  • Tenant isolation
  • Immutable profile versions
  • White-label configuration

Material risk: Direct SIP records unconditionally while disclosure is optional.

03 / Scorecard

Quality first, evidence confidence beside it.

The fixed rubric totals 100 points. Missing and marketing-only evidence is penalized; it is never silently treated as confirmed behavior.

RankCandidateConversation qualityTelephony and call controlContractor workflow fitIntegrations and actionsPost-call artifactsPrivacy and securityOperations and testingMulti-tenant and resellerTotal cost of ownershipTotalConfidence
1 Retell + TJ control plane
Basis

Best combined independent platform evidence and a strong managed telephony/tool substrate. Contractor policy, tenant controls, transfer recovery, voicemail, privacy defaults, and behavioral proof remain incomplete.

Sources: Retell · Retell · Retell · Retell
 21/2512/159/158/107/85/86/75/74/5 77/100 medium-high
2 Synthflow agency / white-label
Basis

Agency-shaped controls, packaged actions, and workflow breadth reduce launch burden. Agreement-specific rates, parent limits, external billing transition, voicemail evidence, and runtime quality cap the score.

Sources: Synthflow · Synthflow · Synthflow · Synthflow
 18/2511/1511/159/107/85/85/76/73/5 75/100 medium
3 Twilio ConversationRelay + OpenAI hybrid
Basis

High conversational ceiling and maximum workflow/control ownership, while reusing TJ knowledge. It also leaves transfer recovery, voicemail, product tenancy, retention, artifacts, and operations with TJ; the current system has no completed-call proof.

Sources: Twilio Conversation Relay · OpenAI Realtime · Twilio · Twilio
 22/2510/158/159/106/85/84/76/74/5 74/100 medium
4 Smith.ai AI + human fallback
Basis

Human escalation and staffed operations create the strongest packaged safety net. AI-only evidence, white-label tenancy, contractor-system depth, and call-based economics limit agency-default fit.

Sources: Smith.ai · Smith.ai · SaaS user · SaaS user
 20/2513/1510/157/106/86/85/72/72/5 71/100 medium
5 LiveKit Agents + TJ product layer
Basis

The strongest integrated open runtime and a credible escape hatch, but still a framework/cloud substrate rather than a receptionist product. TJ would own telephony edges, contractor workflows, governance, tenancy, billing, and support.

Sources: LiveKit Agents · livekit · Gokul JS · Main Branch
 20/259/155/158/105/86/85/76/73/5 67/100 medium
6 Dialzara packaged / partner path
Basis

Clear receptionist packaging and public minute economics make a useful speed-to-market control. Thin independent technical evidence, incomplete partner/API/tenant terms, and vendor-claimed workflow depth require a large confidence penalty.

Sources: Dialzara · Dialzara · Dialzara · SaaS user
 15/2511/1510/156/106/84/84/74/74/5 64/100 low-medium

Rejected strategies

StrategyWhy rejected nowRevisit when
Fully self-host open source first LiveKit/Pipecat and the audited repositories are substrates or demos. TJ would inherit media, telephony, contractor workflows, tenancy, billing, privacy, QA, and 24/7 operations before proving customer value. Managed candidates fail quality, control, or economics tests and TJ has funded platform operations.
Treat the existing Twilio/OpenAI build as production-ready The source proves a narrow controller, not a completed-call receptionist. It lacks structured actions, transfer/REFER, voicemail, bounded hold, fallback, tenant isolation, immutable profiles, and white-label controls. The later protocol passes on a completed TJ-owned product layer.
Opaque packaged product as the universal default Fast setup does not compensate for thin tenant, deterministic action, failure receipt, privacy, and contractor-system evidence. A packaged candidate exposes auditable tenant/action contracts and passes the same protocol.
Human-only answering as the agency core Human fallback is a valuable safety benchmark but call-based economics, limited white-label control, and variable workflow depth constrain scalable resale. High-value or regulated callers justify the cost, or it is used only as escalation.
FSM-native AI as the universal offer Housecall Pro, Jobber, ServiceTitan, and Workiz can be strongest inside their own suites but create customer-stack lock-in and uneven portability. A customer's existing FSM is fixed and its native AI passes the protocol.

04 / Economics

Quality can outweigh cost. Hidden cost cannot.

Representative view: 500 minutes per client, five clients, sold at $300 per client/month. Support assumptions are included; unavailable vendor terms stay unavailable.

Only Retell, Dialzara, and the Twilio/OpenAI hybrid have both a frozen score and comparable numeric planning inputs. Synthflow and Smith.ai stay unplotted because agreement- or call-based terms cannot be converted honestly.

PathMonthly costRevenueGross profitMarginBoundary
Retell + TJ contractor control plane $1,160 $1,500 $340 23.0% $0.16/minute representative landed provider input + $2 number/client + 3 platform-support hours/client at $50/hour; excludes onboarding, taxes, payment processing, compliance, and bespoke integrations.
Dialzara Business Pro $1,313 $1,500 $188 13.0% One plan per client + 0.75 managed-support hour/client at $50/hour; partner wholesale, payment processing, taxes, onboarding, and custom integrations unavailable.
Twilio ConversationRelay + OpenAI hybrid $1,027 $1,500 $473 32.0% One number/client + 3 platform-support hours/client at $50/hour; model allowance is an assumption to replace with measured tokens; excludes recording, storage, transcription, SMS, taxes, incidents, and implementation.
ElevenLabs ElevenAgents Pro $1,308 $1,500 $193 13.0% Separate client workspace/account boundary + 3 platform-support hours/client at $50/hour; pass-through allowance is an assumption; excludes burst usage, taxes, compliance, onboarding, and custom integrations.
Explore all required economics scenarios

0 scenarios shown

PathMinutesClientsSale/clientFixedUsageSupportTotal costRevenueGross profitMargin
Retell + TJ contractor control plane2001At cost$2$32$150$184$184$00.0%
Retell + TJ contractor control plane2001$200$2$32$150$184$200$168.0%
Retell + TJ contractor control plane2001$300$2$32$150$184$300$11639.0%
Retell + TJ contractor control plane2001$400$2$32$150$184$400$21654.0%
Retell + TJ contractor control plane2001$500$2$32$150$184$500$31663.0%
Retell + TJ contractor control plane2005At cost$10$160$750$920$920$00.0%
Retell + TJ contractor control plane2005$200$10$160$750$920$1,000$808.0%
Retell + TJ contractor control plane2005$300$10$160$750$920$1,500$58039.0%
Retell + TJ contractor control plane2005$400$10$160$750$920$2,000$1,08054.0%
Retell + TJ contractor control plane2005$500$10$160$750$920$2,500$1,58063.0%
Retell + TJ contractor control plane20020At cost$40$640$3,000$3,680$3,680$00.0%
Retell + TJ contractor control plane20020$200$40$640$3,000$3,680$4,000$3208.0%
Retell + TJ contractor control plane20020$300$40$640$3,000$3,680$6,000$2,32039.0%
Retell + TJ contractor control plane20020$400$40$640$3,000$3,680$8,000$4,32054.0%
Retell + TJ contractor control plane20020$500$40$640$3,000$3,680$10,000$6,32063.0%
Retell + TJ contractor control plane5001At cost$2$80$150$232$232$00.0%
Retell + TJ contractor control plane5001$200$2$80$150$232$200-$32-16.0%
Retell + TJ contractor control plane5001$300$2$80$150$232$300$6823.0%
Retell + TJ contractor control plane5001$400$2$80$150$232$400$16842.0%
Retell + TJ contractor control plane5001$500$2$80$150$232$500$26854.0%
Retell + TJ contractor control plane5005At cost$10$400$750$1,160$1,160$00.0%
Retell + TJ contractor control plane5005$200$10$400$750$1,160$1,000-$160-16.0%
Retell + TJ contractor control plane5005$300$10$400$750$1,160$1,500$34023.0%
Retell + TJ contractor control plane5005$400$10$400$750$1,160$2,000$84042.0%
Retell + TJ contractor control plane5005$500$10$400$750$1,160$2,500$1,34054.0%
Retell + TJ contractor control plane50020At cost$40$1,600$3,000$4,640$4,640$00.0%
Retell + TJ contractor control plane50020$200$40$1,600$3,000$4,640$4,000-$640-16.0%
Retell + TJ contractor control plane50020$300$40$1,600$3,000$4,640$6,000$1,36023.0%
Retell + TJ contractor control plane50020$400$40$1,600$3,000$4,640$8,000$3,36042.0%
Retell + TJ contractor control plane50020$500$40$1,600$3,000$4,640$10,000$5,36054.0%
Retell + TJ contractor control plane10001At cost$2$160$150$312$312$00.0%
Retell + TJ contractor control plane10001$200$2$160$150$312$200-$112-56.0%
Retell + TJ contractor control plane10001$300$2$160$150$312$300-$12-4.0%
Retell + TJ contractor control plane10001$400$2$160$150$312$400$8822.0%
Retell + TJ contractor control plane10001$500$2$160$150$312$500$18838.0%
Retell + TJ contractor control plane10005At cost$10$800$750$1,560$1,560$00.0%
Retell + TJ contractor control plane10005$200$10$800$750$1,560$1,000-$560-56.0%
Retell + TJ contractor control plane10005$300$10$800$750$1,560$1,500-$60-4.0%
Retell + TJ contractor control plane10005$400$10$800$750$1,560$2,000$44022.0%
Retell + TJ contractor control plane10005$500$10$800$750$1,560$2,500$94038.0%
Retell + TJ contractor control plane100020At cost$40$3,200$3,000$6,240$6,240$00.0%
Retell + TJ contractor control plane100020$200$40$3,200$3,000$6,240$4,000-$2,240-56.0%
Retell + TJ contractor control plane100020$300$40$3,200$3,000$6,240$6,000-$240-4.0%
Retell + TJ contractor control plane100020$400$40$3,200$3,000$6,240$8,000$1,76022.0%
Retell + TJ contractor control plane100020$500$40$3,200$3,000$6,240$10,000$3,76038.0%
Dialzara Business Pro2001At cost$99$0$38$137$137$00.0%
Dialzara Business Pro2001$200$99$0$38$137$200$6432.0%
Dialzara Business Pro2001$300$99$0$38$137$300$16455.0%
Dialzara Business Pro2001$400$99$0$38$137$400$26466.0%
Dialzara Business Pro2001$500$99$0$38$137$500$36473.0%
Dialzara Business Pro2005At cost$495$0$188$683$683$00.0%
Dialzara Business Pro2005$200$495$0$188$683$1,000$31832.0%
Dialzara Business Pro2005$300$495$0$188$683$1,500$81855.0%
Dialzara Business Pro2005$400$495$0$188$683$2,000$1,31866.0%
Dialzara Business Pro2005$500$495$0$188$683$2,500$1,81873.0%
Dialzara Business Pro20020At cost$1,980$0$750$2,730$2,730$00.0%
Dialzara Business Pro20020$200$1,980$0$750$2,730$4,000$1,27032.0%
Dialzara Business Pro20020$300$1,980$0$750$2,730$6,000$3,27055.0%
Dialzara Business Pro20020$400$1,980$0$750$2,730$8,000$5,27066.0%
Dialzara Business Pro20020$500$1,980$0$750$2,730$10,000$7,27073.0%
Dialzara Business Pro5001At cost$99$126$38$263$263$00.0%
Dialzara Business Pro5001$200$99$126$38$263$200-$63-31.0%
Dialzara Business Pro5001$300$99$126$38$263$300$3813.0%
Dialzara Business Pro5001$400$99$126$38$263$400$13834.0%
Dialzara Business Pro5001$500$99$126$38$263$500$23848.0%
Dialzara Business Pro5005At cost$495$630$188$1,313$1,313$00.0%
Dialzara Business Pro5005$200$495$630$188$1,313$1,000-$313-31.0%
Dialzara Business Pro5005$300$495$630$188$1,313$1,500$18813.0%
Dialzara Business Pro5005$400$495$630$188$1,313$2,000$68834.0%
Dialzara Business Pro5005$500$495$630$188$1,313$2,500$1,18848.0%
Dialzara Business Pro50020At cost$1,980$2,520$750$5,250$5,250$00.0%
Dialzara Business Pro50020$200$1,980$2,520$750$5,250$4,000-$1,250-31.0%
Dialzara Business Pro50020$300$1,980$2,520$750$5,250$6,000$75013.0%
Dialzara Business Pro50020$400$1,980$2,520$750$5,250$8,000$2,75034.0%
Dialzara Business Pro50020$500$1,980$2,520$750$5,250$10,000$4,75048.0%
Dialzara Business Pro10001At cost$99$351$38$488$488$00.0%
Dialzara Business Pro10001$200$99$351$38$488$200-$288-144.0%
Dialzara Business Pro10001$300$99$351$38$488$300-$188-62.0%
Dialzara Business Pro10001$400$99$351$38$488$400-$88-22.0%
Dialzara Business Pro10001$500$99$351$38$488$500$133.0%
Dialzara Business Pro10005At cost$495$1,755$188$2,438$2,438$00.0%
Dialzara Business Pro10005$200$495$1,755$188$2,438$1,000-$1,438-144.0%
Dialzara Business Pro10005$300$495$1,755$188$2,438$1,500-$938-62.0%
Dialzara Business Pro10005$400$495$1,755$188$2,438$2,000-$438-22.0%
Dialzara Business Pro10005$500$495$1,755$188$2,438$2,500$633.0%
Dialzara Business Pro100020At cost$1,980$7,020$750$9,750$9,750$00.0%
Dialzara Business Pro100020$200$1,980$7,020$750$9,750$4,000-$5,750-144.0%
Dialzara Business Pro100020$300$1,980$7,020$750$9,750$6,000-$3,750-62.0%
Dialzara Business Pro100020$400$1,980$7,020$750$9,750$8,000-$1,750-22.0%
Dialzara Business Pro100020$500$1,980$7,020$750$9,750$10,000$2503.0%
Twilio ConversationRelay + OpenAI hybrid2001At cost$1$22$150$173$173$00.0%
Twilio ConversationRelay + OpenAI hybrid2001$200$1$22$150$173$200$2714.0%
Twilio ConversationRelay + OpenAI hybrid2001$300$1$22$150$173$300$12742.0%
Twilio ConversationRelay + OpenAI hybrid2001$400$1$22$150$173$400$22757.0%
Twilio ConversationRelay + OpenAI hybrid2001$500$1$22$150$173$500$32765.0%
Twilio ConversationRelay + OpenAI hybrid2005At cost$6$109$750$864$864$00.0%
Twilio ConversationRelay + OpenAI hybrid2005$200$6$109$750$864$1,000$13614.0%
Twilio ConversationRelay + OpenAI hybrid2005$300$6$109$750$864$1,500$63642.0%
Twilio ConversationRelay + OpenAI hybrid2005$400$6$109$750$864$2,000$1,13657.0%
Twilio ConversationRelay + OpenAI hybrid2005$500$6$109$750$864$2,500$1,63665.0%
Twilio ConversationRelay + OpenAI hybrid20020At cost$23$434$3,000$3,457$3,457$00.0%
Twilio ConversationRelay + OpenAI hybrid20020$200$23$434$3,000$3,457$4,000$54314.0%
Twilio ConversationRelay + OpenAI hybrid20020$300$23$434$3,000$3,457$6,000$2,54342.0%
Twilio ConversationRelay + OpenAI hybrid20020$400$23$434$3,000$3,457$8,000$4,54357.0%
Twilio ConversationRelay + OpenAI hybrid20020$500$23$434$3,000$3,457$10,000$6,54365.0%
Twilio ConversationRelay + OpenAI hybrid5001At cost$1$54$150$205$205$00.0%
Twilio ConversationRelay + OpenAI hybrid5001$200$1$54$150$205$200-$5-3.0%
Twilio ConversationRelay + OpenAI hybrid5001$300$1$54$150$205$300$9532.0%
Twilio ConversationRelay + OpenAI hybrid5001$400$1$54$150$205$400$19549.0%
Twilio ConversationRelay + OpenAI hybrid5001$500$1$54$150$205$500$29559.0%
Twilio ConversationRelay + OpenAI hybrid5005At cost$6$271$750$1,027$1,027$00.0%
Twilio ConversationRelay + OpenAI hybrid5005$200$6$271$750$1,027$1,000-$27-3.0%
Twilio ConversationRelay + OpenAI hybrid5005$300$6$271$750$1,027$1,500$47332.0%
Twilio ConversationRelay + OpenAI hybrid5005$400$6$271$750$1,027$2,000$97349.0%
Twilio ConversationRelay + OpenAI hybrid5005$500$6$271$750$1,027$2,500$1,47359.0%
Twilio ConversationRelay + OpenAI hybrid50020At cost$23$1,085$3,000$4,108$4,108$00.0%
Twilio ConversationRelay + OpenAI hybrid50020$200$23$1,085$3,000$4,108$4,000-$108-3.0%
Twilio ConversationRelay + OpenAI hybrid50020$300$23$1,085$3,000$4,108$6,000$1,89232.0%
Twilio ConversationRelay + OpenAI hybrid50020$400$23$1,085$3,000$4,108$8,000$3,89249.0%
Twilio ConversationRelay + OpenAI hybrid50020$500$23$1,085$3,000$4,108$10,000$5,89259.0%
Twilio ConversationRelay + OpenAI hybrid10001At cost$1$109$150$260$260$00.0%
Twilio ConversationRelay + OpenAI hybrid10001$200$1$109$150$260$200-$60-30.0%
Twilio ConversationRelay + OpenAI hybrid10001$300$1$109$150$260$300$4013.0%
Twilio ConversationRelay + OpenAI hybrid10001$400$1$109$150$260$400$14035.0%
Twilio ConversationRelay + OpenAI hybrid10001$500$1$109$150$260$500$24048.0%
Twilio ConversationRelay + OpenAI hybrid10005At cost$6$543$750$1,298$1,298$00.0%
Twilio ConversationRelay + OpenAI hybrid10005$200$6$543$750$1,298$1,000-$298-30.0%
Twilio ConversationRelay + OpenAI hybrid10005$300$6$543$750$1,298$1,500$20213.0%
Twilio ConversationRelay + OpenAI hybrid10005$400$6$543$750$1,298$2,000$70235.0%
Twilio ConversationRelay + OpenAI hybrid10005$500$6$543$750$1,298$2,500$1,20248.0%
Twilio ConversationRelay + OpenAI hybrid100020At cost$23$2,170$3,000$5,193$5,193$00.0%
Twilio ConversationRelay + OpenAI hybrid100020$200$23$2,170$3,000$5,193$4,000-$1,193-30.0%
Twilio ConversationRelay + OpenAI hybrid100020$300$23$2,170$3,000$5,193$6,000$80713.0%
Twilio ConversationRelay + OpenAI hybrid100020$400$23$2,170$3,000$5,193$8,000$2,80735.0%
Twilio ConversationRelay + OpenAI hybrid100020$500$23$2,170$3,000$5,193$10,000$4,80748.0%
ElevenLabs ElevenAgents Pro2001At cost$99$5$150$254$254$00.0%
ElevenLabs ElevenAgents Pro2001$200$99$5$150$254$200-$54-27.0%
ElevenLabs ElevenAgents Pro2001$300$99$5$150$254$300$4615.0%
ElevenLabs ElevenAgents Pro2001$400$99$5$150$254$400$14637.0%
ElevenLabs ElevenAgents Pro2001$500$99$5$150$254$500$24649.0%
ElevenLabs ElevenAgents Pro2005At cost$495$25$750$1,270$1,270$00.0%
ElevenLabs ElevenAgents Pro2005$200$495$25$750$1,270$1,000-$270-27.0%
ElevenLabs ElevenAgents Pro2005$300$495$25$750$1,270$1,500$23015.0%
ElevenLabs ElevenAgents Pro2005$400$495$25$750$1,270$2,000$73037.0%
ElevenLabs ElevenAgents Pro2005$500$495$25$750$1,270$2,500$1,23049.0%
ElevenLabs ElevenAgents Pro20020At cost$1,980$100$3,000$5,080$5,080$00.0%
ElevenLabs ElevenAgents Pro20020$200$1,980$100$3,000$5,080$4,000-$1,080-27.0%
ElevenLabs ElevenAgents Pro20020$300$1,980$100$3,000$5,080$6,000$92015.0%
ElevenLabs ElevenAgents Pro20020$400$1,980$100$3,000$5,080$8,000$2,92037.0%
ElevenLabs ElevenAgents Pro20020$500$1,980$100$3,000$5,080$10,000$4,92049.0%
ElevenLabs ElevenAgents Pro5001At cost$99$13$150$262$262$00.0%
ElevenLabs ElevenAgents Pro5001$200$99$13$150$262$200-$62-31.0%
ElevenLabs ElevenAgents Pro5001$300$99$13$150$262$300$3913.0%
ElevenLabs ElevenAgents Pro5001$400$99$13$150$262$400$13935.0%
ElevenLabs ElevenAgents Pro5001$500$99$13$150$262$500$23948.0%
ElevenLabs ElevenAgents Pro5005At cost$495$63$750$1,308$1,308$00.0%
ElevenLabs ElevenAgents Pro5005$200$495$63$750$1,308$1,000-$308-31.0%
ElevenLabs ElevenAgents Pro5005$300$495$63$750$1,308$1,500$19313.0%
ElevenLabs ElevenAgents Pro5005$400$495$63$750$1,308$2,000$69335.0%
ElevenLabs ElevenAgents Pro5005$500$495$63$750$1,308$2,500$1,19348.0%
ElevenLabs ElevenAgents Pro50020At cost$1,980$250$3,000$5,230$5,230$00.0%
ElevenLabs ElevenAgents Pro50020$200$1,980$250$3,000$5,230$4,000-$1,230-31.0%
ElevenLabs ElevenAgents Pro50020$300$1,980$250$3,000$5,230$6,000$77013.0%
ElevenLabs ElevenAgents Pro50020$400$1,980$250$3,000$5,230$8,000$2,77035.0%
ElevenLabs ElevenAgents Pro50020$500$1,980$250$3,000$5,230$10,000$4,77048.0%
ElevenLabs ElevenAgents Pro10001At cost$99$25$150$274$274$00.0%
ElevenLabs ElevenAgents Pro10001$200$99$25$150$274$200-$74-37.0%
ElevenLabs ElevenAgents Pro10001$300$99$25$150$274$300$269.0%
ElevenLabs ElevenAgents Pro10001$400$99$25$150$274$400$12632.0%
ElevenLabs ElevenAgents Pro10001$500$99$25$150$274$500$22645.0%
ElevenLabs ElevenAgents Pro10005At cost$495$125$750$1,370$1,370$00.0%
ElevenLabs ElevenAgents Pro10005$200$495$125$750$1,370$1,000-$370-37.0%
ElevenLabs ElevenAgents Pro10005$300$495$125$750$1,370$1,500$1309.0%
ElevenLabs ElevenAgents Pro10005$400$495$125$750$1,370$2,000$63032.0%
ElevenLabs ElevenAgents Pro10005$500$495$125$750$1,370$2,500$1,13045.0%
ElevenLabs ElevenAgents Pro100020At cost$1,980$500$3,000$5,480$5,480$00.0%
ElevenLabs ElevenAgents Pro100020$200$1,980$500$3,000$5,480$4,000-$1,480-37.0%
ElevenLabs ElevenAgents Pro100020$300$1,980$500$3,000$5,480$6,000$5209.0%
ElevenLabs ElevenAgents Pro100020$400$1,980$500$3,000$5,480$8,000$2,52032.0%
ElevenLabs ElevenAgents Pro100020$500$1,980$500$3,000$5,480$10,000$4,52045.0%

Sale prices are actual monthly resale prices per client, not markups. At-cost equals modeled cost per client. Support labor is included; public-rate gaps and explicit planning allowances remain visible. Call-, unique-caller-, credit-, interaction-, token-, and quote-based products are not silently converted into minute prices.

05 / Product explorer

54 ways to answer a phone.

Search, filter, and sort the commercial and platform landscape. Links open the canonical first-party product page; contradictions remain visible.

54 products shown

ProductCategoryBest fitWhat it isContradiction or gap
Smith.ai hybrid-managed SMB and high-value callers AI receptionist with live-agent escalation and per-call plans. Rich qualification, scheduling and human fallback; reseller mechanics unavailable.
Goodcall packaged-ai SMB No-code AI phone agent priced by unique callers. Transfer/message/callback flows; retention and reseller rights incomplete.
My AI Front Desk packaged-ai-white-label SMB and agencies Packaged AI workforce with retail and partner paths. Retention and credit/overage language conflict; wholesale starts at a non-guaranteed rate.
Rosie packaged-ai SMB Packaged receptionist with calendar and transfer tiers. Public allowances, but no public overage or managed reseller path.
Dialzara packaged-ai-white-label SMB and agencies Packaged receptionist with public minute tiers and white-label path. Strong public reseller signal; independent behavioral proof still required.
RingCentral AIR suite-standalone SMB and RingCentral customers Standalone or RingEX-adjacent AI receptionist. $39 add-on and $49 standalone contexts must remain separate; bundle overage unavailable.
Ruby human-hybrid SMB Human-first receptionist service with AI enhancements. Strong caller handling but different category and materially higher human-service cost.
Retell managed-developer-platform Builders and product teams Managed phone-agent platform with simulations, transfers, artifacts and provider choice. All-in price depends on voice/model/telephony/add-ons; tenant product layer remains application-owned.
Vapi developer-platform Builders and product teams Developer orchestration platform with tools, transfer, eval and storage controls. Hosting excludes providers; compliance and multi-tenant product controls remain buyer-owned.
Zoom AI Receptionist suite-standalone SMB and Zoom Phone customers Packaged AI receptionist with a standalone path. AI-specific retention and suite/number dependencies require confirmation.
CloudTalk AI Voice Agent suite-addon Sales and support teams AI agent tied to a cloud-phone/contact-center stack. Entry pricing, booking availability, retention and standalone suitability are ambiguous.
Aircall AI Voice Agent suite-standalone SMB and contact-center teams AI voice agent adjacent to Aircall's phone suite. Product-specific consent, retention and current outbound semantics are incomplete.
Abby Connect human-hybrid SMB Human receptionist service with an AI path and public partner link. Partner link is not proof of tenant isolation or white-label rights.
Slang AI vertical-packaged Restaurants Restaurant-native phone agent with reservation and multi-location workflows. Strong vertical fit, weak contractor generality and no public white-label terms.
Synthflow no-code-agency-platform Agencies and builders No-code voice platform with subaccounts and white-label controls. Rates are agreement-specific and in-product Stripe reselling is scheduled to end 2026-09-15.
Housecall Pro CSR AI fsm-native Home-service contractors Housecall Pro-native call/chat handling using business context and availability. Marketing/help disagree on SMS scope; package and independent behavior need verification.
Jobber Receptionist fsm-native Home-service contractors Jobber-native receptionist with published triggers, transfers, transcripts and usage controls. Availability and behavioral proof remain vendor-controlled.
ServiceTitan AI Voice Agents fsm-native Larger home-service contractors ServiceTitan-native booking/dispatch voice automation. Eligibility is package/phone dependent and public total cost is unavailable.
Podium Larry home-services-suite Home-service contractors Home-services AI employee positioned around lead response and dispatch. Integration and dispatch-board claims are vendor-controlled and unpriced.
Workiz Genius Answering fsm-native Service contractors Workiz-native answering, booking and skill/availability dispatch. Older help content warns of omissions/hallucinations; current reliability unproven.
ElevenLabs Contractors platform-template Contractors and builders Contractor intake/urgency template on ElevenAgents. Template is not a packaged FSM product; telephony/model costs and workflow layer remain.
Sameday AI contractor-platform Home-service contractors Contractor voice platform with public FSM, zone and reseller signals. Pricing and independent performance evidence require confirmation.
Dispatcho contractor-packaged Service contractors Packaged agent with minute tiers and emergency routing. Public integration depth is sparse.
Pingwin.ai source unavailable unavailable Unknown Direct public product source unavailable during research. No claims accepted.
Cleo source unavailable contractor-packaged HVAC, plumbing and electrical Contractor-positioned receptionist with Jobber/Housecall Pro claims. Public price and fallback detail unavailable.
Voctiv local-business-packaged Local businesses Generic scheduling and qualification phone agent. Older $29 claim is stale and unconfirmed.
Rossy managed-voice-agent Service trades and agencies Managed agent with trade coverage and reseller signal. Template-copy contradictions lower confidence.
HomeServices AI agency-bundle Home-service contractors Website/SEO/CRM/receptionist bundle. Not a standalone receptionist comparator.
Wellgrow Lyra contractor-packaged Home-service contractors Published service-area checks, emergency routing and multi-location tiers. Extensive outcome claims are vendor-controlled; independent proof required.
FusionCaller service-business-packaged Service businesses Inbound agent with public setup/monthly pricing. Affiliate availability is not white-label proof.
AfterCall source unavailable unavailable Unknown Direct product source returned unavailable during research. No claims accepted.
Bland managed-platform Builders and enterprise teams Managed phone-agent platform with self-serve, VPC and testing paths. Telephony is separate; retention/transfer/security and latency claims remain vendor-controlled.
ElevenLabs ElevenAgents managed-agent-platform Builders and businesses Voice-agent platform with telephony, transfer, voicemail, experimentation and retention controls. LLM and telephony are separately billed; no behavioral proof.
Twilio Conversation Relay cpaas-transport Builders Managed STT/TTS/session transport over programmable telephony. Application owns LLM/orchestration, tools, consent, tenancy, artifacts, failure policy and QA.
OpenAI Realtime realtime-model-api Builders Realtime speech/model API with SIP, VAD, functions and MCP. Not a receptionist operating system; PSTN product layer and phone-specific QA are external.
LiveKit Agents runtime-media-cloud Builders RTC/SIP voice-agent runtime with interruption, tools and tests. Contractor workflow, tenant, consent, artifacts, billing and operations remain application-owned.
Daily Bots managed-realtime-infrastructure Builders Realtime bot infrastructure with WebRTC/PSTN and provider integrations. SIP and full handoff semantics are incomplete; product layer remains.
Pipecat open-source-framework Builders Composable multimodal/voice pipeline framework. Telephony policy, retention, tenancy, artifacts and production operations remain adopter-owned.
Voiceflow visual-cx-platform Agencies and CX teams Visual agent platform with partner workspaces and phone workflows. Voice dollar rates and low-level call semantics are unclear.
Telnyx Voice AI carrier-orchestration Builders Carrier plus managed voice orchestration. $0.05 engine headline excludes LLM and often telephony; artifact/eval depth incomplete.
Deepgram Voice Agent API managed-agent-api Builders Single-WebSocket speech/agent API. PSTN bridge, transfer, artifacts and current exact all-in rate remain external or unclear.
Hume EVI expressive-speech-api Builders Expressive speech-to-speech interaction API. No native public PSTN/SIP receptionist stack.
Google Conversational Agents managed-conversation-platform Enterprise builders Managed deterministic/generative conversation and telephony tooling. Production phone behavior and total cost are integration-dependent.
Nextiva XBert suite-standalone SMB Standalone-capable AI receptionist/agent path in a business-communications vendor. Interaction-based pricing and suite dependencies need careful normalization.
Dialpad AI suite-addon Business communications customers AI voice/contact-center capabilities within Dialpad. Autonomous receptionist suitability and suite dependency require separation.
GoTo Connect AI Receptionist suite-addon SMB phone customers AI receptionist tied to GoTo Connect. Parent phone-suite and entitlement dependency.
Genesys Cloud AI contact-center-addon Enterprise contact centers Enterprise AI voice/contact-center stack. Not a lightweight standalone receptionist; services and quote burden material.
Five9 GenAI contact-center-addon Enterprise contact centers AI capabilities in Five9's contact-center suite. Suite/services dependency and quote-based economics.
NICE CXone AI contact-center-addon Enterprise contact centers AI capabilities in a broad contact-center suite. Standalone suitability, services and quote burden.
Talkdesk AI contact-center-addon Enterprise contact centers AI agents/features in Talkdesk's suite. Suite dependency and public retention/reseller gaps.
Cisco Webex AI Agent contact-center-addon Enterprise communications AI agent capability tied to Webex/contact-center tooling. Connector/professional-service and suite dependencies.
Vonage AI suite-platform Builders and enterprises Communications APIs/AI Studio path. Total phone-product layer and economics depend on broader Vonage services.
Microsoft Teams Phone Agent preview-suite-addon Microsoft 365 customers Phone-agent capability in public preview. Preview latency/voice limits; not production proof.
8x8 AI contact-center-addon Business and enterprise communications AI/contact-center capability tied to 8x8. Suite dependency, quote terms and receptionist-specific evidence gaps.

06 / Open source

Exactly 12 deep source audits.

Reviewed SHAs pin the evidence. A sample or framework is not promoted to a production receptionist.

RepositoryReviewed SHALicenseArchitectureProduction gap
bolna-ai/bolna 8e5a04cedd3a MIT Broad telephony/orchestration substrate with SIP/BYOT, interruption, DTMF, tools and Redis. Hosted UI/API closed; tenant, consent, policy and production proof missing.
vocodedev/vocode-core e054c33a7278 MIT Historical phone-agent framework with Twilio, state, actions and transfer abstractions. Stale core and hosted/API split; modern security, tenancy, retention and operations unresolved.
pipecat-ai/pipecat d6583ee89bc9 BSD-2-Clause Highly composable voice pipeline with Twilio/Daily adapters, turn strategies, flows and workers. Carrier policy, transfer/voicemail/failure semantics, artifacts, tenancy and operations remain.
TEN-framework/ten-framework 028494298183 Apache-derived restricted license Active extension/runtime ecosystem with RTC/SIP examples. License restrictions are a go/no-go diligence item; public sample lacks production security/state.
livekit/agents 88014ce3033a Apache-2.0 plus bundled model terms Integrated RTC/SIP runtime with turn handling, AMD, tools, warm transfer, traces and eval hooks. Contractor workflow, tenant, consent, artifacts, billing and operations remain application-owned.
bentoml/BentoVoiceAgent 83ea2ab54d71 Unavailable root repository license Minimal Twilio/Pipecat/BentoCloud demonstration. Nearly the entire secure multi-client product/control plane is missing.
NVIDIA-AI-Blueprints/nemotron-voice-agent cb996534eb57 BSD-2-Clause plus vendor model/container terms GPU/Pipecat voice, turn-taking and evaluation substrate. No PSTN/SIP or receptionist product layer; substantial GPU/vendor burden.
Azure-Samples/call-center-voice-agent-accelerator c26e8a003c32 MIT Managed PSTN/provider bridge to Azure Voice Live with operational scaffolding. Accelerator with process-local state and no complete receptionist workflow or durable artifacts.
Azure-Samples/art-voice-agent-accelerator 8b29308f3140 MIT Application-shaped ACS/SIP, transfer, DTMF, handoff, state, metrics and eval reference. Demonstration disclaimer plus auth/CORS, PII logging, Redis TLS, tenancy and production-proof gaps.
openai/openai-voice-agent-sdk-sample 6cead90cee27 MIT Official browser/FastAPI Agents SDK starting point. No PSTN, continuous duplex phone control, durable state, tenancy or production controls.
twilio/media-streams 09d208b2ce2f Unavailable root license Official media/WebSocket demonstration repository. Transport demos only; the entire receptionist runtime and product layer remains.
langfuse/langfuse 24c4f687d514 MIT core plus commercially licensed Enterprise areas Observability, evaluation and retention substrate. No voice/telephony runtime; Cloud versus heavy self-host and EE gates require review.

07 / Independent evidence

269 qualifying opinion URLs.

The artifact preserves 292 canonical secondary URLs; 269 remain in the qualifying independent count after disclosed affiliation exclusions. One review aggregate is one URL.

269 qualifying items shown

Publisher/sourceFamilySubjectsThemeSentimentConfidenceGap
AI_Agent_Reviews user Reddit Retell, Vapi, Synthflow, Bland, retell, vapi, synthflow, bland voice quality, latency, interruptions, conversion mix M self-reported metrics not independently verified
RingCentral user Reddit RingCentral AIR, ringcentral-air missing email notifications, Calendly limits, admin monitoring neg H single customer report
AgentsOfAI user Reddit Vapi, Synthflow, Bland, Retell, vapi, synthflow, bland, retell developer control, latency, no-code limits mix M small test sample
RingCentral user Reddit RingCentral AIR, ringcentral-air routing flexibility, repeated recording notice mix H configuration-specific
RingCentral user Reddit RingCentral AIR, ringcentral-air cancellation delay, recurring billing neg H account-specific terms unavailable
AI_Agents user Reddit Bland, Synthflow, Retell, bland, synthflow, retell voice naturalness, interruptions, conversion mix M self-reported campaign results
AI_Agent_Reviews user Reddit Retell, Vapi, Synthflow, Bland, retell, vapi, synthflow, bland production readiness, compliance, analytics mix M claims not externally audited
aiagents user Reddit Vapi, Bland, Retell, Synthflow, vapi, bland, retell, synthflow Vapi immaturity, Retell bugs, Bland access, price mix M small bake-off
AI_Agent_Reviews user Reddit Retell, Vapi, Synthflow, Bland, retell, vapi, synthflow, bland latency, summaries, integrations, human escalation pos M self-reported results
Best_Ai_Agents user Reddit Retell, Vapi, Synthflow, PolyAI, retell, vapi, synthflow naturalness, calendar integration, enterprise burden pos M comparative preference not benchmark
AI_Agents user Reddit Retell, Vapi, Bland, retell, vapi, bland handoff, booking, summaries, prompt iteration mix H detailed retrospective, unverified metrics
aiagents user Reddit Retell, DIY stack, retell latency, memory, glue code, scalability mix M single builder account
AgentsOfAI user Reddit Retell, custom stack, retell full-duplex, interruptions, analytics pos M early testing
AI_Agents contractor Reddit Bland, Vapi, Retell, bland, vapi, retell requirements, warm transfer, loops, client adoption neg H contractor experience, possible agency interest
AI_Agent_Reviews user Reddit Retell, Bland, Vapi, retell, bland, vapi naturalness, latency, interruptions pos M no quantified call set
AI_Agents user Reddit Retell, retell architecture, ASR/TTS, state, telephony pos M technical inference mixed with experience
AIToolTesting user Reddit Retell, retell latency, interruptions, context loss mix M heavy-load stability unresolved
aiagents user Reddit Retell, Vapi, Twilio, retell, vapi, twilio-conversation-relay latency, memory, prompt tuning mix M comparison lacks controlled method
AI_Agent_Reviews user Reddit Retell, retell voice quality, context, workflow pos M visible date not precise
Best_Ai_Agents user Reddit Retell, retell turn-taking, SIP, KB, support pos M no independent measurement
Best_Ai_Agents user Reddit Retell, retell voice quality, integration, reliability pos M date shown approximately
Best_Ai_Agents user Reddit Retell, retell stability, support, workflow glue pos M date not precisely visible
aiagents user Reddit Vapi, vapi observability, loops, escalation mix H single implementation
aiagents user Reddit Bland, Retell, Vapi, Synthflow, bland, retell, vapi, synthflow workflow completion, engineering burden mix M benchmark method not published
AI_Agents user Reddit Vapi, Bland, Retell, vapi, bland, retell opaque billing, barge-in tuning, migration neg H author has competing project
TextToSpeech user Reddit ElevenLabs, Cartesia, Murf, elevenlabs-elevenagents TTFB, p95 latency, cost, code mixing mix H not a controlled benchmark
ElevenLabs user Reddit ElevenLabs Agents, elevenlabs-elevenagents call ending, preview/live mismatch, GUI bugs neg H single user account
ElevenLabs user Reddit ElevenLabs, elevenlabs-elevenagents documentation, UX, outages, billing neg H general product and agent evidence mixed
ArtificialInteligence user Reddit ElevenLabs, elevenlabs-elevenagents voice quality, regeneration cost, drift mix M content-creation emphasis
ElevenLabs user Reddit ElevenLabs Agents, elevenlabs-elevenagents webhooks, summaries, billing unit, privacy mix H reseller affiliation disclosed
ElevenLabs user Reddit ElevenLabs Agents, elevenlabs-elevenagents language, latency, startup delay neg H date displayed approximately
VoiceAutomationAI user Reddit ElevenLabs Enterprise, elevenlabs-elevenagents voice quality, pricing, number handling mix M small discussion
ElevenLabs user Reddit ElevenLabs Agents, elevenlabs-elevenagents latency, model choice, turn-taking mix H advice thread, not independent benchmark
ElevenLabs user Reddit ElevenLabs, Bland, Alta, Synthflow, elevenlabs-elevenagents, bland, synthflow robotic feel, pauses, CRM need mix H single operator report
OpenAI user Reddit OpenAI Realtime, openai-realtime tool calling, browsing, context, architecture pos M early prototype
aiagents user Reddit OpenAI Realtime, Vapi, openai-realtime, vapi latency, rate limits, prompts pos M strong preference, little measurement
ArtificialInteligence user Reddit OpenAI Realtime 2, openai-realtime tools, prompt parity, guardrails mix M preview/new-model evidence
Twilio user Reddit Twilio, OpenAI, Vapi, Retell, twilio-conversation-relay, openai-realtime, vapi, retell outbound scheduling, state, cost, interruptions mix M personal project rather than business deployment
twilio user Reddit Twilio, OpenAI Realtime, twilio-conversation-relay, openai-realtime barge-in, code-enforced guardrails, maintenance pos H self-reported implementation
twilio user Reddit Twilio ConversationRelay, Gemini, twilio-conversation-relay WebSocket, context, concurrent sessions pos H tutorial-style user post
twilio user Reddit Twilio ConversationRelay, twilio-conversation-relay latency, interruption, turn-taking, handoff pos M community post references vendor materials
twilio user Reddit Twilio ConversationRelay, twilio-conversation-relay low latency, WebSocket, OpenAI pos M date shown approximately
twilio user Reddit Twilio ConversationRelay, Gemini, twilio-conversation-relay STT/TTS streaming, context pos M date shown approximately
AIReceptionists user Reddit AI receptionist category setup burden, action capability, monitoring mix M products not consistently named
AI_Agents user Reddit AI receptionist category customer trust, loops, configuration mix M product-specific attribution weak
AIReceptionists user Reddit AI receptionist category message taking, automation limits mix M product names not fixed
AIReceptionists user Reddit Goodcall, Nextiva, Marblism, goodcall after-hours, booking, human fallback mix M multiple products, no controlled comparison
AIReceptionists user Reddit Rosie, rosie answering quality, information calls pos M product-specific detail limited
aiagents user Reddit Rosie, rosie human-like voice, intake pos M date not precisely shown
Entrepreneur user Reddit Rosie, rosie SMS notifications, booking, lead handling pos M mixed product mentions
aiagents user Reddit Goodcall, goodcall UI, integrations, appointment use mix M client/consultant perspective
AI_Agents user Reddit Goodcall, AI receptionist, goodcall 24/7 answering, routing, UI mix M product attribution mixed
smallbusiness user Reddit My AI Front Desk, my-ai-front-desk booking, cost, missed-call recovery neutral M prospective rather than deployed
automation user Reddit My AI Front Desk, my-ai-front-desk AI quality, cancellation, Calendly dependency neg H contract/affiliation claims unverified
GoHighLevel user Reddit My AI Front Desk, my-ai-front-desk setup, workflows, white-label, quality mix M mixed owner/promotional replies
AIReceptionists user Reddit Rosie, My AI Front Desk, Goodcall, rosie, my-ai-front-desk, goodcall price, call volume, feature fit mix M comparison not a controlled test
SaaS user Reddit Dialzara, Voice.ai, Smith.ai, dialzara, smith-ai white-glove setup, comparative testing mix M product outcome not fully disclosed
AiForSmallBusiness user Reddit Dialzara, My AI Front Desk, Bland, dialzara, my-ai-front-desk, bland basic answering, integrations, setup mix M broad AI-employee discussion
SaaS user Reddit Dialzara, Smith.ai, Sonant, dialzara, smith-ai pricing, multilingual, setup, quality mix M author discloses co-founder interest
NHS user Reddit AI receptionists trust, medical errors, accessibility neg H provider not identified
BetterOffline user Reddit AI receptionists missed messages, wrong reminders, privacy neg H provider identities unavailable
AI solobusinesses user Reddit Dialzara, Vapi, GoHighLevel, dialzara, vapi plug-and-play versus developer platform mix M comparison page referenced rather than fully tested
AI_Agents user Reddit CloudTalk, Retell, Vapi, retell, vapi setup, CRM, human handoff mix M CloudTalk-specific comments may be promotional
aiagents user Reddit voice-agent stacks listening layer, STT, latency, evals neg M product attribution broad
independent AI builder YouTube Retell, retell voice quality, setup, workflow testing pos M creator methodology not fully visible
independent AI builder YouTube Synthflow, synthflow interruption handling, latency, integrations mix M creator methodology not fully visible
CX Foundation YouTube RingCentral AIR, ringcentral-air voice/SMS, routing, lead capture, booking pos M case-study claims not audited
All About AI YouTube OpenAI Realtime, openai-realtime latency, prompts, realtime behavior mix M early model version
Shweta Lodha YouTube OpenAI Realtime, openai-realtime setup, WebRTC, implementation pos M tutorial rather than production
Southern Neat/CX creator YouTube RingCentral AIR, ringcentral-air onboarding, configuration, observed features pos M new product exposure
independent builder YouTube OpenAI, Twilio voice stack, openai-realtime, twilio-conversation-relay voice quality, turn-taking, build burden mix M limited metadata
Dograh builder YouTube Vapi alternative, voice stack, vapi barge-in, telephony, KB, tools mix M competitor bias disclosed
independent OpenAI builder YouTube OpenAI Realtime, openai-realtime voice quality, latency, tool use pos M short demo
independent ElevenLabs user YouTube ElevenLabs Agents, elevenlabs-elevenagents multilingual turn-taking, transcription error mix M linked from HN discussion
independent developer YouTube OpenAI Realtime, openai-realtime tools, WebRTC, implementation pos M early version
independent voice creator YouTube ElevenLabs, elevenlabs-elevenagents naturalness, cloning, latency pos L primarily synthesis, limited agent evidence
Vapi founder and commenters Hacker News Vapi, vapi reliability, latency, tuning, uptime mix M vendor claims mixed with independent comments
independent Gmail voice-app builder Hacker News Retell, Vapi, retell, vapi privacy, retention, tool use, utility pos H specific implementation details
LiveKit/OpenAI builders and commenters Hacker News OpenAI Realtime, openai-realtime packet loss, WebRTC, latency mix H technical discussion
Vapi builder Hacker News Vapi, PlayHT, vapi latency, API, voice cloning pos M self-reported build
Vapi demo users Hacker News Vapi, ElevenLabs, vapi, elevenlabs-elevenagents voice recognition, metering, recording deletion mix M older platform version
ElevenLabs users/commenters Hacker News ElevenLabs Agents, elevenlabs-elevenagents transcription error, cost, tool use mix M vendor employee participated
OpenAI audio users Hacker News OpenAI Realtime, ElevenLabs, openai-realtime, elevenlabs-elevenagents voice cutoffs, VAD, production readiness neg H older API version but concrete failure report
ElevenLabs users Hacker News ElevenLabs, elevenlabs-elevenagents naturalness, latency, cloning pos M older synthesis version
OpenAI Realtime builder Hacker News OpenAI Realtime, openai-realtime tools, WebRTC, deployment pos M short comment thread
developer binkrassdufass OpenAI Community OpenAI Realtime, openai-realtime progressive latency, context compaction neg H specific model/version
developers OpenAI Community OpenAI Realtime, openai-realtime structured telephony, cold-call workflow, tools mix M new model discussion
developer comparison OpenAI Community OpenAI, ElevenLabs, Deepgram, openai-realtime, elevenlabs-elevenagents subsecond latency, stack architecture mix M date approximate
Retell user Retell Community Retell, retell gibberish, loops, reliability neg H support forum, provider unresolved
Retell user Retell Community Retell, retell minute accounting, support access neg H billing claim unresolved
AI Automation Society users Skool community Vapi, Retell, vapi, retell latency, function calling, CRM booking, reseller portals mix M commercial claims unverified
Brendan AI Community Skool community Vapi, Retell, vapi, retell KB determinism, quoting, agency use pos M small discussion
Google Voice user Google Voice Community Smith.ai, smith-ai routing, greeting, setup failure neg M provider/account configuration unclear
automation user Pabbly forum Vapi, Retell, vapi, retell appointment reminders, integration need neutral L prospective rather than deployed
Ryan Whitton Independent blog Goodcall, goodcall setup time, calendar, GBP, complexity ceiling pos M commercial comparison basis
WorkflowStack AI Independent blog Goodcall, Rosie, Smith.ai, goodcall, rosie, smith-ai setup, voice quality, price, niche fit mix M self-reported production basis
Altin Sallaku Independent blog Rosie, rosie voice, booking, human fallback absence, price pos M affiliate link and positive verdict
Smarter Clicks Independent blog Rosie, rosie setup effort, booking, accuracy pos M price conflicts with other sources
Appscribed Independent blog My AI Front Desk, my-ai-front-desk price, setup, call quality, trust mix M rebrand and pricing may drift
SchedulingKit Independent blog My AI Front Desk, my-ai-front-desk ease, booking, complex-call limits mix M pricing appears stale/conflicting
AI and Realtors Independent blog My AI Front Desk, my-ai-front-desk setup, real-estate workflow, booking mix M testing method not visible
Stellar Independent blog Smith.ai, Goodcall, smith-ai, goodcall human fallback, setup, integrations mix M not fully hands-on
Katherine Stone/CX Foundation Independent blog RingCentral AIR, ringcentral-air call routing, booking, lead capture, pricing pos M demo line and case claims not independently audited
Front Desk Review Independent blog Dialzara, dialzara price, call economics, positioning mix M aggregation methodology not fully visible
Sonant Competitor blog Dialzara, dialzara policy/quote calls, documentation, alternatives neg M direct competitive bias
Wave Runner Agency blog Retell, Synthflow, Vapi, retell, synthflow, vapi agency setup, white-label, cost, reporting mix H platform owner/competitor bias
Ryan Whitton Independent blog Retell, Vapi, Bland, Synthflow, retell, vapi, bland, synthflow latency, cost, use-case fit mix M method detail limited
StackBriefly Independent blog Vapi, Retell, ElevenLabs, Bland, Synthflow, vapi, retell, elevenlabs-elevenagents, bland, synthflow developer fit, voice quality, cost mix M review methodology limited
Omid Saffari Independent blog Goodcall, Rosie, Smith.ai, Synthflow, Retell, Bland, goodcall, rosie, smith-ai, synthflow, retell, bland billing unit, fit, price mix M comparative pricing may be stale
Hermes Agency blog Vapi, Retell, Synthflow, vapi, retell, synthflow latency, agency deployment, platform fit mix M agency bias
Ringvox Agency blog Synthflow, Vapi, Retell, Bland, synthflow, vapi, retell, bland European SMB, GDPR, voice quality mix M competitor bias
Wyse Tools Independent blog Retell, retell ops discipline, privacy, interruptions pos M method details limited
Lucia Franzese Substack Vapi, vapi engineering burden, flexibility, voice quality mix H early model/product version
Ryan Whitton Independent blog Retell, retell latency, voice quality, integrations, price pos M internal testing not independently audited
GrowwStacks Independent blog Synthflow, synthflow voice quality, interruption, integration, price mix M method not independently audited
VoiceAI Guide Independent blog Retell, retell setup, text simulation, latency, cost pos M methodology incomplete
Top AI Voice Agents Independent blog Twilio ConversationRelay, twilio-conversation-relay telephony, orchestration, pricing, integration burden mix M review basis not fully visible
TechRadar Tech publication Bland AI, bland voice naturalness, looping, disclosure mix H single public call
TechRadar Pro Tech publication Zoom AI Receptionist standalone deployment, phone-system dependency mix M limited hands-on detail
Intelligent Conversations: AI in Dentistry Apple Podcasts AI receptionist, Yobi dental workflows, scheduling, HIPAA, training mix M vendor pick disclosed
APIs You Won’t Hate Podcast site Vapi, vapi API design, latency, onboarding, voice-agent architecture mix M founder interview, not independent product test
Forward Deployed Spotify Vapi, Retell, Smallest, Daily, vapi, retell latency, tool calls, QA, HIPAA, retention, turn-taking mix H discussion not controlled product test
MLOps.community Spotify Voice-agent systems cascade failures, authentication, latency, QA mix H not product-specific
The Wealthy Contractor Spotify Custom voice agent 30 simultaneous calls, transcript review, missed-call recovery pos H custom stack, not product-specific
Open.cx Technical blog OpenAI Realtime, openai-realtime full-duplex architecture, tool calls, latency mix M tutorial rather than independent benchmark
Dev Note Technical blog OpenAI Realtime, openai-realtime function calling, turn detection, error handling mix M measured claims not independently audited
TECHSY Technical blog OpenAI Realtime, openai-realtime tool calls, interruption, latency pos M tutorial methodology limited
Sam Eddy Technical blog Twilio ConversationRelay, ElevenLabs, Deepgram, Claude, twilio-conversation-relay, elevenlabs-elevenagents multi-hop latency, transport, outbound/inbound mix H specific architecture, no public call dataset
Mostafa Ibrahim Technical blog Twilio ConversationRelay, twilio-conversation-relay tools, WebSocket, action execution pos M tutorial rather than production report
WeAreDevelopers / speakers Developer conference Twilio ConversationRelay, twilio-conversation-relay STT/TTS abstraction, latency, infrastructure burden mix M speaker affiliation disclosed
independent authors arXiv Realtime voice-agent stack, ElevenLabs, elevenlabs-elevenagents P50 time-to-first-audio, cloud versus self-hosted mix H paper not customer-production evidence
independent authors arXiv OpenAI, Gemini, Qwen realtime systems, openai-realtime tone comprehension, turn-taking, semantic listening mix H benchmark scope differs from receptionist workflows
independent authors OpenReview Full-duplex voice agents latency, responsiveness, interruptions, tool use mix H not vendor-specific receptionist deployment
NerdyNav Technical blog ElevenLabs, elevenlabs-elevenagents voice quality, STT, conversational-agent suitability pos M content-creator emphasis
Tom’s Guide Tech publication ElevenLabs Agents, elevenlabs-elevenagents setup time, turn-taking, phone number, pricing pos H single test and account-dependent
Tom’s Guide Tech publication ChatGPT, ElevenLabs, elevenlabs-elevenagents voice naturalness, tone, playback workflow pos M not phone-agent focused
MorphLLM editorial desk editorial-comparison Vapi, Retell, Bland, Synthflow, vapi, retell, bland, synthflow landed-cost and architecture tradeoffs mixed medium no public call logs or tenant-isolation test
Hand On Web editorial-comparison Retell, Vapi, Synthflow, retell, vapi, synthflow regional telephony and latency mixed medium sample size and call recordings unavailable
Loic Bachellerie editorial-comparison Vapi, Bland, Retell, vapi, bland, retell developer experience and customization mixed medium no independently reproduced benchmarks
Nextelligentia editorial-comparison Retell, Vapi, Bland, Synthflow, retell, vapi, bland, synthflow deployment burden and cost mixed low testing protocol and affiliation unclear
Sacesta editorial-comparison Retell, Vapi, Synthflow, Bland, retell, vapi, synthflow, bland feature fit and usability mixed low hands-on depth and data unavailable
FuturePicker editorial-comparison Retell, Vapi, Synthflow, Bland, retell, vapi, synthflow, bland pricing, setup and use-case fit mixed low advertorial incentives not disclosed
Get Daily Toolbox editorial-comparison Retell, Vapi, Bland, Synthflow, ElevenLabs, retell, vapi, bland, synthflow, elevenlabs-elevenagents pricing and broad fit positive/mixed low primarily desk research
ZGLG editorial-comparison Vapi, Retell, Bland, Synthflow, ElevenLabs, vapi, retell, bland, synthflow, elevenlabs-elevenagents feature breadth and deployment mixed low no raw hands-on evidence
Contractor ToolStack editorial-comparison My AI Front Desk, my-ai-front-desk review credibility and trust signals negative/mixed medium limited public independent review volume
Smarter Clicks AI editorial-comparison Rosie, rosie voice quality and call handling positive/mixed medium call sample and failure cases unavailable
Appscribed editorial-comparison Ruby, ruby human fallback, cost and target vertical positive/mixed medium AI-specific behavior less applicable because Ruby is human-led
Fit Small Business editorial-comparison Ruby, ruby pricing, human answering and setup positive/mixed high not an AI-agent implementation test
ConsumerAffairs editorial-comparison Ruby, ruby transfer reliability and support mixed/negative medium reviewer identity and representativeness vary
TRTC editorial-comparison Ruby, Smith.ai, ruby, smith-ai human-versus-AI tradeoff mixed low commercial relationship unclear
Claudessa editorial-comparison Ruby, ruby customization and handoff mixed low no reproducible call evidence
Scored Tools hands-on-review Vapi, vapi flexibility and engineering burden mixed low test basis not fully documented
RedoYou hands-on-review Vapi, vapi developer fit and setup mixed low limited reproducibility
VoiceRank hands-on-review Vapi, vapi pipeline ownership and customization positive/mixed low review methodology unclear
DroidCrunch hands-on-review Vapi, vapi setup and pricing mixed medium no call recordings or benchmark data
SmartRepl hands-on-review Vapi, vapi developer experience and stack control mixed low hands-on detail limited
Tom’s Guide hands-on-review ElevenLabs, elevenlabs-elevenagents setup, interruption handling and naturalness positive high consumer-scale test, not enterprise tenancy
Tested Media hands-on-review Vapi, vapi flexibility, setup and cost mixed medium underlying provider pass-throughs not fully evidenced
AICX Stack hands-on-review Vapi, vapi latency, integrations and deployment mixed low vendor latency claims not independently measured
Tested Media benchmark-research AI voice agents, human receptionists detectability and call outcomes positive/mixed medium study design and sample recruitment unavailable
Tested Media benchmark-research Retell, Synthflow, Vapi, Bland, retell, synthflow, vapi, bland reliability and no-code versus technical setup positive/mixed medium raw call corpus unavailable
SipPulse benchmark-research voice-agent stacks observability and production metrics mixed medium platform-specific results not all exposed
AlphaRun benchmark-research voice-agent platforms real-call performance and workflow fit mixed medium test scripts and raw results incomplete
OpenBenchmarks benchmark-research Telnyx, Bland, ElevenLabs, Retell, Vapi, bland, elevenlabs-elevenagents, retell, vapi latency measurement and benchmark validity mixed high not all vendor configurations public
Himanshu Rawat benchmark-research voice-agent stacks latency, VAD and barge-in mixed medium no vendor-by-vendor results
DestiLabs benchmark-research voice-agent platforms latency, cost and containment mixed medium client and platform configuration details unavailable
Guideflow benchmark-research voice assistants setup and workflow breadth mixed low benchmark data not exposed
Nick Tikhonov benchmark-research custom voice-agent stack latency, streaming and interruption positive/mixed medium custom stack not directly comparable to managed APIs
Chatbotscape operator-writeup Voiceflow, voiceflow knowledge, tools and handoff mixed medium test corpus and scoring rubric unavailable
AI Demos operator-writeup Voiceflow, voiceflow builder usability and integrations mixed low limited evidence on telephony behavior
Gokul JS operator-writeup LiveKit, livekit-agents-product implementation burden and pipeline control mixed medium no production scale or failure-rate data
Tried by Humans operator-writeup Voiceflow, voiceflow setup and integration depth mixed medium test methodology not fully reproducible
Cadence operator-writeup Daily, Pipecat, pipecat-product deployment burden and framework choice mixed low commercial relationship unclear
Main Branch operator-writeup LiveKit, livekit-agents-product agent framework and RAG positive/mixed medium limited telephony evidence
The Neuron operator-writeup LiveKit, livekit-agents-product production readiness and tooling mixed medium interview rather than controlled test
Techsy developer-stack-comparison OpenAI Realtime, Twilio, openai-realtime, twilio-conversation-relay implementation burden and interruption handling mixed medium OpenAI product facts excluded unless corroborated by official sources
Skywork AI developer-stack-comparison OpenAI Realtime, openai-realtime latency, tools and implementation speed positive/mixed medium external opinion only, not authoritative product specification
YourStory developer-stack-comparison OpenAI Realtime, openai-realtime tools and handoff architecture mixed low demo evidence is not production evidence
Latent Space developer-stack-comparison OpenAI Realtime, Pipecat, openai-realtime, pipecat-product architecture and framework tradeoffs mixed medium older source; current API behavior may have changed
APIScout developer-stack-comparison OpenAI Realtime, Gemini, Vapi, Retell, Twilio, Deepgram, ElevenLabs, openai-realtime, vapi, retell, twilio-conversation-relay, elevenlabs-elevenagents managed versus lower-level API burden mixed medium pricing and feature claims need current official corroboration
Ry Walker Research developer-stack-comparison OpenAI Realtime, Gemini, ElevenLabs, Cartesia, Vapi, Retell, LiveKit, Pipecat, Deepgram, openai-realtime, elevenlabs-elevenagents, vapi, retell, livekit-agents-product, pipecat-product pass-through pricing and API ownership mixed high pricing is time-sensitive and configuration-dependent
Amjid Ali developer-stack-comparison Retell, Vapi, LiveKit, retell, vapi, livekit-agents-product developer control and voice quality mixed medium sample and test harness unavailable
Ikki developer-stack-comparison ElevenLabs, Vapi, Retell, elevenlabs-elevenagents, vapi, retell voice quality, tools and orchestration mixed low no reproducible call evidence
Akash Maurya developer-stack-comparison Twilio, LiveKit, Vapi, Bland, twilio-conversation-relay, livekit-agents-product, vapi, bland stack ownership, cost and compliance mixed medium no independent billing export or call corpus
CelloIP developer-stack-comparison Vapi, Retell, LiveKit, Bland, Twilio, Deepgram, OpenAI, ElevenLabs, vapi, retell, livekit-agents-product, bland, twilio-conversation-relay, openai-realtime, elevenlabs-elevenagents setup time and component ownership mixed low marketing influence possible
independent video author developer-stack-comparison LiveKit, ElevenLabs, Vapi, Retell, Synthflow, GoHighLevel, livekit-agents-product, elevenlabs-elevenagents, vapi, retell, synthflow setup and support mixed low video methodology and sponsorship unclear
Presenc AI benchmark-research Vapi, Retell, Synthflow, Bland, OpenAI, Cartesia, ElevenLabs, vapi, retell, synthflow, bland, openai-realtime, elevenlabs-elevenagents latency and deployment maturity mixed medium measurement method and confidence intervals unavailable
AIAgentRank benchmark-research Vapi, Retell, Bland, vapi, retell, bland price and operational fit mixed low ranking rubric not fully transparent
Clearcall benchmark-research AI receptionist platforms regional voice quality and GDPR concerns mixed medium results are described separately from methodology
Awesome Agents benchmark-research ElevenLabs, Vapi, Retell, Bland, Play.ai, elevenlabs-elevenagents, vapi, retell, bland latency, pricing and use cases mixed low raw measurements unavailable
Silverthread Labs benchmark-research Vapi, Retell, Bland, ElevenLabs, vapi, retell, bland, elevenlabs-elevenagents latency and architecture mixed medium benchmark environment unavailable
TechRadar editorial-comparison voice-agent platforms CRM integration, compliance and fallback mixed high not a product-specific hands-on test
TechRadar editorial-comparison Quo, Sona small-business phone UX and artifacts positive/mixed high limited evidence on high-volume telephony
TechRadar editorial-comparison Ooma Office routing, pricing and setup mixed high AI receptionist depth less than specialized agents
TechRadar editorial-comparison Grasshopper telephony fallback and portability mixed high not a full conversational AI agent
TechRadar editorial-comparison RingCentral, Five9, Genesys, Talkdesk enterprise routing and analytics mixed high list article, not controlled tests for every product
TechRadar editorial-comparison Aircall call quality, integrations and analytics mixed high AI agent behavior not the primary focus
TechRadar editorial-comparison Intermedia, RingCentral, Nextiva, Zoom enterprise phone cost and complexity mixed high comparative claims are not controlled tests
TechRadar editorial-comparison RingCentral, Nextiva, Zoom, Ooma telephony coverage and integration mixed high AI features vary by plan and region
TechRadar editorial-comparison RingCentral, Nextiva, Zoom, Aircall carrier coverage and feature depth mixed medium older review date for some products
TechRadar editorial-comparison Bland, bland naturalness, looping and disclosure mixed high celebrity demo is not business workflow evidence
G2 G2-user-review Retell, retell voice quality, latency and cost positive/mixed high aggregate review bias and vendor responses
G2 G2-user-review Vapi, vapi flexibility and setup burden positive/mixed high small review base relative to Retell
G2 G2-user-review Bland, bland setup and feature breadth mixed medium small sample and seller-managed profile
G2 G2-user-review Synthflow, synthflow no-code setup and pricing positive/mixed high AI-generated summary may overgeneralize
G2 G2-user-review Smith.ai, smith-ai screening, transfer and caller acceptance positive/mixed high small sample and hybrid human service
G2 G2-user-review Smith.ai, smith-ai human fallback and consistency mixed high different product from Smith AI Receptionist
G2 G2-user-review RingCentral AIR, ringcentral-air AI receptionist quality and setup mixed medium small review sample
G2 G2-user-review Telnyx Voice AI latency and connectivity positive/mixed medium small sample and possible vendor-invited reviews
G2 G2-user-review CallHippo AI Voice Agent language support and reliability mixed medium review volume and test depth limited
G2 G2-user-review Leaping AI automation and integrations positive/mixed medium some product claims are vendor profile text
G2 G2-user-review Voiceflow, voiceflow builder, tools and observability positive/mixed high voice telephony evidence thinner than chat evidence
G2 G2-user-review AgentVoice automation and CRM actions positive/mixed low small sample and possible vendor influence
G2 G2-user-review Voicing AI tools and analytics positive/mixed low all reviews positive and sample tiny
G2 G2-user-review CallmAi voice automation and languages mixed low limited review volume
G2 G2-user-review VoiceGenie accuracy and call automation positive/mixed low small review corpus
G2 G2-user-review Vodex automation and deployment positive/mixed low small sample and sparse details
G2 G2-user-review Conversica lead qualification and follow-up positive/mixed medium voice-specific evidence limited
G2 G2-user-review Thoughtly setup, voice quality and learning curve positive/mixed high review text may include incentivized submissions
G2 G2-user-review Pyto outbound sales and follow-up positive/mixed medium small and specialized user base
G2 G2-user-review Slang AI restaurant calls and customization positive/mixed high some guest and incentivized reviews
G2 G2-user-review ElevenLabs, elevenlabs-elevenagents voice quality and editing positive/mixed high mostly speech-generation evidence rather than telephony
G2 G2-user-review Ooma Office routing and pricing positive/mixed medium pricing page mixes product and review content
G2 G2-user-review AnswerAide appointment handling and reliability mixed medium only two reviews and older dates
TrustRadius TrustRadius-user-review Genesys Cloud CX enterprise voice, bot and analytics positive/mixed high summary aggregates many heterogeneous deployments
TrustRadius TrustRadius-user-review Rasa customization and self-hosting burden mixed medium voice-specific sample not isolated
TrustRadius TrustRadius-user-review LivePerson handoff and omnichannel support mixed medium product scope broader than voice agent
Capterra Capterra-user-review Smith.ai, smith-ai setup, integrations and support positive/mixed high product page combines chat and receptionist offerings
Capterra Capterra-user-review Ruby, ruby human answering and consistency positive/mixed high not an AI-agent product
Capterra Capterra-user-review Botpress builder usability and learning curve mixed high voice implementation evidence limited
Capterra Capterra-user-review Kore.ai enterprise deployment and support positive/mixed medium voice details are not isolated from broader bot use
Capterra Capterra-user-review Cognigy low-code tooling and support positive/mixed high review dates span several years
Capterra Capterra-user-review Genesys Cloud CX enterprise routing and complexity mixed high AI-specific observations are uneven
Capterra Capterra-user-review Talkdesk reporting and integration friction mixed high not all evidence concerns AI features
Capterra Capterra-user-review Five9 outbound operations and analytics positive/mixed high AI-agent behavior is not consistently reported
Capterra Capterra-user-review Nextiva support and administration mixed/negative high review sentiment varies by account configuration
Capterra Capterra-user-review Dialpad transcription, latency and analytics mixed high incentivized and non-incentivized samples mixed
Capterra Capterra-user-review Aircall call quality and AI artifacts mixed high product scope broader than voice agent
Capterra Capterra-user-review Vonage Contact Center routing and enterprise operations mixed medium AI-specific evidence sparse
Capterra Capterra-user-review Ooma Office routing and integrations mixed high not a dedicated conversational AI platform
Capterra Capterra-user-review Zoom Phone call quality and admin UX positive/mixed high AI receptionist evidence limited
Capterra Capterra-user-review Vonage Business Communications routing and reliability mixed high AI-agent behavior not isolated
Capterra Capterra-user-review 8x8 Work reliability and administration mixed medium AI-specific evidence sparse
Capterra Capterra-user-review CloudTalk call quality and billing negative/mixed high single severe review may not generalize
Capterra Capterra-user-review JustCall AI agent, reporting and support mixed/negative high review sentiment and incentives vary
Capterra Capterra-user-review RingCentral RingEX telephony UX and remote deployment positive/mixed medium AIR evidence not isolated
Capterra Capterra-user-review RingCentral Contact Center routing, analytics and administration mixed medium AIR-specific behavior sparse
Software Advice SoftwareAdvice-user-review SAP Conversational AI builder and integrations positive/mixed high review is older than page update
Software Advice SoftwareAdvice-user-review Mava naturalness and integration positive/mixed medium older evidence and chat-first scope
Software Advice SoftwareAdvice-user-review Intelswift accuracy, inbox consolidation and support positive/mixed medium voice-specific evidence limited
Software Advice SoftwareAdvice-user-review Amelia enterprise automation and deployment mixed low review detail is sparse
Software Advice SoftwareAdvice-user-review fonio.ai voice quality and accuracy positive low review volume and methodology limited
Trustpilot Trustpilot-user-review AI Receptionist setup, Spanish handling and spam filtering positive medium small sample and likely customer-selection bias
Trustpilot Trustpilot-user-review Smith.ai, smith-ai customization, reliability and support mixed high company replies and heterogeneous products
Trustpilot Trustpilot-user-review ReceptionHQ human service quality and billing mixed medium mostly human receptionist evidence
Trustpilot Trustpilot-user-review Newo voice quality and booking positive low all visible reviews positive
Trustpilot Trustpilot-user-review Clara answering quality and support positive/mixed medium review volume and product configuration unclear
Trustpilot Trustpilot-user-review Voice.ai voice quality, billing and cancellation negative/mixed high consumer voice changer rather than phone-agent API
Trustpilot Trustpilot-user-review Dialora voice quality and support positive/mixed low small and potentially incentivized-looking corpus
Trustpilot Trustpilot-user-review CloneVoice voice quality and pricing mixed medium speech-generation rather than telephony agent
Trustpilot Trustpilot-user-review Answering Service Care transfer and reliability mixed/negative medium human answering service rather than AI
Trustpilot Trustpilot-user-review Centerfy AI support and voice quality positive low small sample and no detailed test protocol
Trustpilot Trustpilot-user-review Nedzo agency deployment and customization positive low small sample and promotional language
Trustpilot Trustpilot-user-review Posh answering quality and support positive/mixed medium primarily human service evidence
Trustpilot Trustpilot-user-review Nextiva AI receptionist reliability and support negative medium single page and product version unclear
Trustpilot Trustpilot-user-review Speechify voice quality and billing mixed medium consumer TTS rather than developer voice-agent API
Trustpilot Trustpilot-user-review Dialzara, dialzara naturalness, booking, transcripts and setup positive medium all visible sentiment positive and vendor support is intertwined
SchedulingKit hands-on-review Dialzara, dialzara setup, features and support mixed medium test script and support correspondence unavailable
Contractor ToolStack hands-on-review Dialzara, dialzara voice choice, setup and pricing mixed medium limited independent customer evidence
Open the 88-row primary, engineering, pricing, and source-audit ledger
Publisher/sourceFamilySubjectsThemeLabelConfidenceGap
Smith.ai first-party-productsmith-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorRich qualification, scheduling and human fallback; reseller mechanics unavailable.
Goodcall first-party-productgoodcallDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorTransfer/message/callback flows; retention and reseller rights incomplete.
My AI Front Desk first-party-productmy-ai-front-deskDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorRetention and credit/overage language conflict; wholesale starts at a non-guaranteed rate.
Rosie first-party-productrosieDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorPublic allowances, but no public overage or managed reseller path.
Dialzara first-party-productdialzaraDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorStrong public reseller signal; independent behavioral proof still required.
RingCentral AIR first-party-productringcentral-airDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behavior$39 add-on and $49 standalone contexts must remain separate; bundle overage unavailable.
Ruby first-party-productrubyDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorStrong caller handling but different category and materially higher human-service cost.
Retell first-party-productretellDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorAll-in price depends on voice/model/telephony/add-ons; tenant product layer remains application-owned.
Vapi first-party-productvapiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorHosting excludes providers; compliance and multi-tenant product controls remain buyer-owned.
Zoom AI Receptionist first-party-productzoom-ai-receptionistDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorAI-specific retention and suite/number dependencies require confirmation.
CloudTalk AI Voice Agent first-party-productcloudtalk-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorEntry pricing, booking availability, retention and standalone suitability are ambiguous.
Aircall AI Voice Agent first-party-productaircall-ai-voice-agentDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorProduct-specific consent, retention and current outbound semantics are incomplete.
Abby Connect first-party-productabby-connectDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorPartner link is not proof of tenant isolation or white-label rights.
Slang AI first-party-productslang-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorStrong vertical fit, weak contractor generality and no public white-label terms.
Synthflow first-party-productsynthflowDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorRates are agreement-specific and in-product Stripe reselling is scheduled to end 2026-09-15.
Housecall Pro CSR AI first-party-producthousecall-pro-csr-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorMarketing/help disagree on SMS scope; package and independent behavior need verification.
Jobber Receptionist first-party-productjobber-receptionistDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorAvailability and behavioral proof remain vendor-controlled.
ServiceTitan AI Voice Agents first-party-productservicetitan-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorEligibility is package/phone dependent and public total cost is unavailable.
Podium Larry first-party-productpodium-larryDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorIntegration and dispatch-board claims are vendor-controlled and unpriced.
Workiz Genius Answering first-party-productworkiz-genius-answeringDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorOlder help content warns of omissions/hallucinations; current reliability unproven.
ElevenLabs Contractors first-party-productelevenlabs-contractorsDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorTemplate is not a packaged FSM product; telephony/model costs and workflow layer remain.
Sameday AI first-party-productsameday-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorPricing and independent performance evidence require confirmation.
Dispatcho first-party-productdispatchoDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorPublic integration depth is sparse.
Voctiv first-party-productvoctivDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorOlder $29 claim is stale and unconfirmed.
Rossy first-party-productrossyDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorTemplate-copy contradictions lower confidence.
HomeServices AI first-party-producthomeservices-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorNot a standalone receptionist comparator.
Wellgrow Lyra first-party-productwellgrow-lyraDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorExtensive outcome claims are vendor-controlled; independent proof required.
FusionCaller first-party-productfusioncallerDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorAffiliate availability is not white-label proof.
Bland first-party-productblandDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorTelephony is separate; retention/transfer/security and latency claims remain vendor-controlled.
ElevenLabs ElevenAgents first-party-productelevenlabs-elevenagentsDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorLLM and telephony are separately billed; no behavioral proof.
Twilio Conversation Relay first-party-producttwilio-conversation-relayDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorApplication owns LLM/orchestration, tools, consent, tenancy, artifacts, failure policy and QA.
OpenAI Realtime first-party-productopenai-realtimeDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorNot a receptionist operating system; PSTN product layer and phone-specific QA are external.
LiveKit Agents first-party-productlivekit-agents-productDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorContractor workflow, tenant, consent, artifacts, billing and operations remain application-owned.
Daily Bots first-party-productdaily-botsDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorSIP and full handoff semantics are incomplete; product layer remains.
Pipecat first-party-productpipecat-productDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorTelephony policy, retention, tenancy, artifacts and production operations remain adopter-owned.
Voiceflow first-party-productvoiceflowDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorVoice dollar rates and low-level call semantics are unclear.
Telnyx Voice AI first-party-producttelnyx-voice-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behavior$0.05 engine headline excludes LLM and often telephony; artifact/eval depth incomplete.
Deepgram Voice Agent API first-party-productdeepgram-voice-agentDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorPSTN bridge, transfer, artifacts and current exact all-in rate remain external or unclear.
Hume EVI first-party-producthume-eviDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorNo native public PSTN/SIP receptionist stack.
Google Conversational Agents first-party-productgoogle-conversational-agentsDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorProduction phone behavior and total cost are integration-dependent.
Nextiva XBert first-party-productnextiva-xbertDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorInteraction-based pricing and suite dependencies need careful normalization.
Dialpad AI first-party-productdialpad-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorAutonomous receptionist suitability and suite dependency require separation.
GoTo Connect AI Receptionist first-party-productgoto-ai-receptionistDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorParent phone-suite and entitlement dependency.
Genesys Cloud AI first-party-productgenesys-cloud-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorNot a lightweight standalone receptionist; services and quote burden material.
Five9 GenAI first-party-productfive9-genaiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorSuite/services dependency and quote-based economics.
NICE CXone AI first-party-productnice-cxone-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorStandalone suitability, services and quote burden.
Talkdesk AI first-party-producttalkdesk-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorSuite dependency and public retention/reseller gaps.
Cisco Webex AI Agent first-party-productcisco-webex-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorConnector/professional-service and suite dependencies.
Vonage AI first-party-productvonage-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorTotal phone-product layer and economics depend on broader Vonage services.
Microsoft Teams Phone Agent first-party-productteams-phone-agentDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorPreview latency/voice limits; not production proof.
8x8 AI first-party-producteight-by-eight-aiDocumented product positioning, features, and commercial availabilityvendor-claimedmedium for documented presence; unavailable for runtime behaviorSuite dependency, quote terms and receptionist-specific evidence gaps.
bolna-ai GitHubbolna-ai/bolnaExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorHosted UI/API closed; tenant, consent, policy and production proof missing.
vocodedev GitHubvocodedev/vocode-coreExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorStale core and hosted/API split; modern security, tenancy, retention and operations unresolved.
pipecat-ai GitHubpipecat-ai/pipecatExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorCarrier policy, transfer/voicemail/failure semantics, artifacts, tenancy and operations remain.
TEN-framework GitHubTEN-framework/ten-frameworkExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorLicense restrictions are a go/no-go diligence item; public sample lacks production security/state.
livekit GitHublivekit/agentsExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorContractor workflow, tenant, consent, artifacts, billing and operations remain application-owned.
bentoml GitHubbentoml/BentoVoiceAgentExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorNearly the entire secure multi-client product/control plane is missing.
NVIDIA-AI-Blueprints GitHubNVIDIA-AI-Blueprints/nemotron-voice-agentExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorNo PSTN/SIP or receptionist product layer; substantial GPU/vendor burden.
Azure-Samples GitHubAzure-Samples/call-center-voice-agent-acceleratorExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorAccelerator with process-local state and no complete receptionist workflow or durable artifacts.
Azure-Samples GitHubAzure-Samples/art-voice-agent-acceleratorExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorDemonstration disclaimer plus auth/CORS, PII logging, Redis TLS, tenancy and production-proof gaps.
openai GitHubopenai/openai-voice-agent-sdk-sampleExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorNo PSTN, continuous duplex phone control, durable state, tenancy or production controls.
twilio GitHubtwilio/media-streamsExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorTransport demos only; the entire receptionist runtime and product layer remains.
langfuse GitHublangfuse/langfuseExact-SHA source, architecture, and license auditverifiedhigh for source presence at reviewed SHA; none for runtime behaviorNo voice/telephony runtime; Cloud versus heavy self-host and EE gates require review.
Retell first-party-documentationretellCurrent pricing range, components, concurrency, and number costsverifiedhigh for documented contract or research finding; unavailable for target runtimeComponent choices prevent one universal all-in rate.
Retell first-party-documentationretellDocumented transfer primitiveverifiedhigh for documented contract or research finding; unavailable for target runtimeWarm/cold terminal behavior still requires testing.
Retell first-party-documentationretellHTTP function/action primitiveverifiedhigh for documented contract or research finding; unavailable for target runtimeContractor authorization, idempotency, and rollback remain TJ-owned.
Retell first-party-documentationretellCall event and artifact webhook contractverifiedhigh for documented contract or research finding; unavailable for target runtimeDelivery reconciliation and retention policy remain application work.
Retell first-party-documentationretellConfigurable data-retention controlsverifiedhigh for documented contract or research finding; unavailable for target runtimeDefault/plan behavior must be frozen for the selected account.
Synthflow first-party-documentationsynthflowUsage and billing mechanicsverifiedhigh for documented contract or research finding; unavailable for target runtimeExact rate is account/agreement-specific.
Synthflow first-party-documentationsynthflowSubaccount permission modelverifiedhigh for documented contract or research finding; unavailable for target runtimeParent-limit and tenant-isolation behavior requires contract/runtime verification.
Synthflow first-party-documentationsynthflowAgency and white-label controlsvendor-claimedmedium; vendor claim onlyIn-product Stripe reselling is documented for removal on 2026-09-15.
Synthflow first-party-documentationsynthflowTransfer modes, timeouts, and retriesverifiedhigh for documented contract or research finding; unavailable for target runtimeVoicemail and terminal-state behavior remain incomplete publicly.
Synthflow first-party-documentationsynthflowDocumented action typesverifiedhigh for documented contract or research finding; unavailable for target runtimeClosed-loop contractor write correctness is unproven.
Smith.ai first-party-documentationsmith-aiIntegration catalogvendor-claimedmedium; vendor claim onlyCatalog presence does not prove bidirectional workflow correctness.
Dialzara first-party-documentationdialzaraPlans, included minutes, overage, and package capabilitiesvendor-claimedmedium; vendor claim onlyCentral parser unavailable; partner/API terms remain incomplete.
Dialzara first-party-documentationdialzaraPackaged receptionist feature claimsvendor-claimedmedium; vendor claim onlyPublic API/auth/tenant contract unavailable.
Twilio first-party-documentationtwilio-conversation-relayConversationRelay transport and session contractverifiedhigh for documented contract or research finding; unavailable for target runtimeApplication owns business logic, recovery, actions, and policy.
Twilio first-party-documentationtwilio-conversation-relayInterruption, handoff, and WebSocket message contractverifiedhigh for documented contract or research finding; unavailable for target runtimeUnexpected disconnect ends session; recovery requires a new Connect flow.
Twilio first-party-documentationtwilio-conversation-relayConversationRelay usage pricingverifiedhigh for documented contract or research finding; unavailable for target runtimeVoice, model, storage, support, and taxes remain additional.
OpenAI first-party-documentationopenai-realtimeRealtime session, audio, and tool primitivesverifiedhigh for documented contract or research finding; unavailable for target runtimeModel primitives do not make a complete receptionist.
OpenAI first-party-documentationopenai-realtimeSIP accept, monitor, refer, and hangup primitivesverifiedhigh for documented contract or research finding; unavailable for target runtimeApplication policy and terminal recovery remain required.
OpenAI first-party-documentationopenai-realtimeServer and semantic VAD controlsverifiedhigh for documented contract or research finding; unavailable for target runtimeVendor thresholds are not acceptance-test proof.
Housecall Pro first-party-documentationhousecall-pro-csr-aiContractor-native CSR AI setup and workflowsvendor-claimedmedium; vendor claim onlyAPI/plan gates and behavioral correctness remain untested.
Jobber first-party-documentationjobber-receptionistContractor-native receptionist workflowsvendor-claimedmedium; vendor claim onlyTransfer, voicemail, and bidirectional booking evidence is incomplete.
ServiceTitan first-party-documentationservicetitan-aiFSM-native AI conversation and booking strategiesvendor-claimedmedium; vendor claim onlySuite/access requirements and runtime behavior remain untested.
Workiz first-party-documentationworkiz-genius-answeringGenius Answering capabilities and limitsvendor-claimedmedium; vendor claim onlyMonthly conversation ceiling can stop AI answering.
τ-Voice authors academic-researchvoice-agent-categoryRealistic voice-agent task and interaction benchmarkverifiedhigh for documented contract or research finding; unavailable for target runtimeBenchmark is not the contractor PSTN-to-artifact lifecycle.
DeVries et al. academic-researchvoice-agent-categoryDeaf and hard-of-hearing access to voice assistantsverifiedhigh for documented contract or research finding; unavailable for target runtimeNot a phone-receptionist product comparison.

08 / Later test plan

Pre-registered calls. Zero results.

Every scenario is explicitly not run. The plan compares the five finalists under identical carrier, audio, workflow, privacy, and failure conditions.

01
not_run

Normal qualification and booking

Expected: One normalized lead and one booking; lifecycle ends terminal_success.

Gate: At least 95% required fields correct; exactly one lead and booking; zero duplicate effects; frozen latency limits met.

Protocol details

Preconditions: Clean synthetic profile; caller provides service, address, urgency, availability, contact details, and confirms one slot.

Caller script: Provide the fixed caller turns in order and confirm the proposed slot once.

Telemetry: Turn P50/P95/P99, required-field extraction, tool attempts, idempotency key, booking/lead IDs, artifact manifest.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

02
not_run

Out-of-area decline

Expected: Explain the boundary and offer only the approved alternative; no booking or unauthorized mutation.

Gate: 100% correct decline; zero false booking, transfer, or follow-up.

Protocol details

Preconditions: Synthetic address outside the configured service polygon.

Caller script: Request an appointment after stating the out-of-area address.

Telemetry: Geofence decision, policy branch, tool count, terminal state.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

03
not_run

Emergency escalation

Expected: Apply frozen emergency policy, approved safety language, and synthetic escalation or safe decline.

Gate: 100% policy-correct handling; no ordinary booking; any unsafe advice or misroute is a P0 stop.

Protocol details

Preconditions: Synthetic gas-smell, sparking, or flooding fixture; no real emergency or number.

Caller script: State the emergency symptom and ask for immediate help.

Telemetry: Classifier, policy version, target, transfer state, terminal state, disclosure.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

04
not_run

After-hours behavior

Expected: Execute the configured emergency, callback, voicemail, or safe-decline branch.

Gate: 100% correct branch; no business-hours promise or unauthorized booking.

Protocol details

Preconditions: Frozen after-hours timestamp and policy.

Caller script: Request routine service, then present the emergency variant.

Telemetry: Effective hours/time zone, branch, mutation, terminal state.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

05
not_run

New or existing customer ambiguity

Expected: Clarify before selection; never merge, overwrite, or expose another profile.

Gate: At least 95% correct clarification; zero unauthorized merge, overwrite, or disclosure.

Protocol details

Preconditions: Zero, one, and multiple matching synthetic profiles.

Caller script: Say 'I have used you before' with an ambiguous name/address.

Telemetry: Candidate matches, clarification turn, profile/version, mutation attempts.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

06
not_run

Trade classification ambiguity

Expected: Ask one bounded clarification and route only after confirmation.

Gate: At least 95% correct classification; zero action before confirmation.

Protocol details

Preconditions: One request plausibly maps to two configured trades.

Caller script: Describe the ambiguous symptom without naming the trade.

Telemetry: Intent candidates, clarification, chosen service, tool payload, terminal state.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

07
not_run

Structured lead capture and correction

Expected: Keep only final confirmed values and create one lead/callback task.

Gate: At least 95% final-field accuracy; one mutation; no stale value.

Protocol details

Preconditions: Synthetic contact details and writable sandbox.

Caller script: Give fields out of order; correct address and preferred callback time.

Telemetry: Field revisions, confirmation, idempotency key, final payload hash.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

08
not_run

Interruption and barge-in

Expected: Cancel/truncate audio, preserve caller content, and queue no post-cancel speech.

Gate: At least 95% accepted; stop P95 at most 300 ms; residual P95 at most 250 ms; zero queued audio after cancel.

Protocol details

Preconditions: Frozen long agent response and two interruption points.

Caller script: Interrupt with 'Wait—Tuesday works,' then interrupt during a tool-related response.

Telemetry: VAD/EOT, interruption acceptance, cancellation IDs, stop latency, residual audio, repetitions.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

09
not_run

Noise and accent robustness

Expected: Preserve confirmed intent/fields; reprompt rather than invent low-confidence data.

Gate: At least 90% required-field correctness; zero irreversible action from unconfirmed data.

Protocol details

Preconditions: Pre-registered accent and controlled clean/noise/network profiles.

Caller script: Repeat the identical qualification script under every condition.

Telemetry: Condition labels, ASR/intent, reprompts, EOT distributions, action attempts.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

10
not_run

Silence and trailing-off speech

Expected: Distinguish pause from EOT; bounded reprompt; no booking from incomplete data.

Gate: Zero incomplete booking; separately reported false/late EOT; no unbounded wait.

Protocol details

Preconditions: Frozen pause durations and incomplete-detail variant.

Caller script: Pause mid-request, trail off, resume, then stop before required details.

Telemetry: Silence duration, false/late EOT, reprompts, P50/P95/P99, tool attempts.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

11
not_run

DTMF

Expected: Map each supported digit exactly once and use defined malformed fallback.

Gate: 100% mapping; no duplicate/missing digit; 100% defined fallback.

Protocol details

Preconditions: Candidate-supported DTMF path and malformed-input fallback.

Caller script: Send 1, 2, 0, #, then an unsupported sequence.

Telemetry: Direction, codec/stream mode, digit IDs/count, malformed branch.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

12
not_run

Transfer answered

Expected: handoff_pending to human_connected to terminal_success; one leg and complete context.

Gate: 100% terminal status; no silent drop/duplicate bridge; at least 95% required context.

Protocol details

Preconditions: Synthetic human target answers and accepts context.

Caller script: Ask for a human and remain on the line through handoff.

Telemetry: Protocol/leg, request/answer times, target disposition, context hash, terminal state.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

13
not_run

Transfer rejected or no-answer

Expected: Explicit voicemail, callback, safe decline, or return-to-caller fallback.

Gate: One terminal outcome; no silent drop/loop; fallback within hold bound.

Protocol details

Preconditions: Synthetic target rejects or does not answer.

Caller script: Ask for transfer and follow the offered fallback.

Telemetry: Attempts, reject/no-answer reason, hold time, fallback, artifact.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

14
not_run

Bounded hold timeout

Expected: Disclose wait, enforce bound, and execute fallback.

Gate: 100% hold at most 30 seconds; no silence/loop; explicit fallback.

Protocol details

Preconditions: Synthetic unresolved queue and 30-second caller-visible hold bound.

Caller script: Accept hold and remain silent until timeout.

Telemetry: Hold start/stop/disclosure, elapsed time, queue, fallback, terminal state.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

15
not_run

Human, voicemail, and unknown detection

Expected: Classify human/voicemail/unknown and preserve unknown when confidence is insufficient.

Gate: Complete matrix; unknown at most 10% where claimed; no human drop or unsafe misclassification.

Protocol details

Preconditions: Labeled human, voicemail, silence, noise, and ambiguous destinations.

Caller script: Present each answer mode using the frozen corpus.

Telemetry: Full confusion matrix, TP/TN/FP/FN, unknown rate, branch outcome.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

16
not_run

Spam or robocall handling

Expected: Apply spam policy with no lead, booking, transfer, or outbound action.

Gate: Zero unauthorized side effect; one explicit safe terminal state.

Protocol details

Preconditions: Synthetic prerecorded, repeat, invalid-profile, and noise patterns.

Caller script: Play each spam fixture without conversational completion.

Telemetry: Classifier, policy branch, mutation count, terminal state, suppression reason.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

17
not_run

Tool timeout, retry, and idempotency

Expected: Policy-bounded retry, caller update, one mutation or safe fallback.

Gate: Zero duplicate effects; at most 10-second caller wait before fallback; terminal state.

Protocol details

Preconditions: Calendar/CRM delay, timeout, and duplicate callback with one idempotency key.

Caller script: Attempt a booking while the synthetic tool faults.

Telemetry: Attempt IDs, timeout/retry, key, callback correlation, side-effect count.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

18
not_run

Carrier, media, or model failure

Expected: Map fault to caller-safe fallback or terminal_failure; never claim success.

Gate: 100% safe terminal mapping; zero unclassified hangs or post-failure effects.

Protocol details

Preconditions: One injected carrier disconnect, media fault, model timeout, cancellation failure, or app exception per run.

Caller script: Proceed through a standard intake until the frozen injection point.

Telemetry: Fault ID, provider/leg state, retries, cancellation, terminal state, artifact status.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

19
not_run

AI disclosure and consent refusal

Expected: Truthful disclosure; honor refusal and continue only where policy permits.

Gate: 100% disclosure/refusal compliance; zero prohibited artifact.

Protocol details

Preconditions: Frozen disclosure, recording, transcript, and retention policy.

Caller script: Ask if it is AI and refuse recording/transcription.

Telemetry: Disclosure, consent, recording/transcription/retention state, terminal status.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

20
not_run

Opt-out and do-not-contact

Expected: Stop prohibited follow-up and write one correctly scoped suppression record only when authorized.

Gate: 100% suppression compliance; zero outbound action after opt-out.

Protocol details

Preconditions: Scoped synthetic suppression registry.

Caller script: Say 'do not call, text, or contact me again.'

Telemetry: Phrase, scope, suppression version, outbound count, final state.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

21
not_run

Artifact deletion and redaction

Expected: Delete/redact content; retain only non-content receipt/tombstone.

Gate: Zero retained raw PII; complete receipt; no alternate-path retrieval.

Protocol details

Preconditions: Synthetic PII in transcript, recording, tool payload, and summary.

Caller script: Request deletion/redaction after the call.

Telemetry: Before/after manifest, redaction, deletion time, denial proof, tombstone hash.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

22
not_run

Artifact completeness and async finalization

Expected: Finalize only when every artifact state is completed, unavailable, or failed.

Gate: 100% required fields or explicit unavailable; no socket-close success; idempotent reconciliation.

Protocol details

Preconditions: Call with qualification, tool action, handoff/fallback, and delayed callbacks.

Caller script: End the call before recording/transcript callbacks complete.

Telemetry: Correlated call/session/turn/tool/transfer/artifact IDs and callback states.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

23
not_run

Tenant and configuration-version isolation

Expected: Use immutable profile snapshot and access only correct tenant/version data.

Gate: Zero cross-tenant/version leakage; every artifact has correct snapshot/hash.

Protocol details

Preconditions: Synthetic tenants A/B and versions V1/V2 with distinct canaries, service areas, calendars, prompts, targets.

Caller script: Repeat identical phrases across all tenant/version combinations.

Telemetry: Tenant/profile/version, config hash, canaries, tool scopes, artifact access checks.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

24
not_run

Quote, callback, and follow-up consent

Expected: Obtain explicit consent; create exactly one task; safe fallback on failure.

Gate: 100% consent; exactly one authorized task; zero action after refusal/timeout.

Protocol details

Preconditions: Synthetic quote/callback tool and scoped consent policy.

Caller script: Request estimate/callback, confirm time, then authorize one follow-up.

Telemetry: Consent, scope, payload, idempotency key, task count, delivery/fallback.

Cleanup: Close every call leg; reconcile callbacks; delete synthetic calendar/CRM/lead/quote/callback/voicemail records; apply deletion/redaction policy; retain only redacted metrics, hashes, and deletion receipts.

09 / Methods

What this report proves—and what it does not.

Reconciliation: 302 worker rows − 10 cross-lane duplicates = 292 preserved canonical URLs; 23 disclosed affiliated/vendor context rows are excluded, leaving 269 qualifying independent URLs.

Saturation: Reached after worker lanes did not saturate. The coordinator continued with materially distinct consumer/vertical, academic, accessibility, consent/liability, procurement, and incident/fallback searches. The final two families added neither a product nor a material decision theme.

Final no-add families: coordinator-procurement-and-rfp-controls; coordinator-production-incident-and-fallback.

  • No vendor account, trial, call, deployment, SDK or behavioral benchmark was executed.
  • Public documentation proves documented capability, not completed-call quality or production parity.

Research window: Current public evidence reviewed through 2026-08-19. Access snapshot: 2026-08-19. No vendor account, trial, call, deployment, or commercial/open-source runtime test was executed.