US BUYER’S GUIDE · 6 PRODUCTS

AI voice agents

Can an agent complete a real phone task and hand off safely when it cannot?

Compare configurable voice agents by deployment, real business actions, handoff, operations and the complete cost of a working call.

Research desk · Oct 20, 2026 · Methodology

YOUR SHORTLIST STARTS HERE

Compare the options.

Sort by

Highest documented score first. Fit adds up the 28 feature checks below, including AI, across documented plans. Entry price does not include every capability. Unknown evidence appears as a range.

Product / good fit for	Pricing context	Documented fit	Next step
Retell AI Teams building production phone agents who want a packaged voice stack and measurable testing workflow.	Usage basedper configured conversation minutePay as you go	80–99/100 documented fit 81% evidence coverage	Full research →
ElevenLabs Agents Teams building branded voice experiences across phone, web or mobile with a configurable tool layer.	$6.00per monthStarter	74–100/100 documented fit 74% evidence coverage	Full research →
Vapi Engineering teams that want to control models, tools, deployment and application behavior.	$0.05per minute hosting onlyBuild	71–100/100 documented fit 71% evidence coverage	Full research →
Synthflow Organizations buying a supported production voice rollout with defined commercial and operational terms.	$30,000.00per year, starting contract · billed annuallyEnterprise	66–100/100 documented fit 66% evidence coverage	Full research →
respond.io B2C teams handling sustained inbound conversation volume across messaging channels.	$79.00per month · billed annuallyStarter	47–100/100 documented fit 47% evidence coverage	Full research →
Bland AI Teams that want explicit conversation paths and a published bundled AI minute rate.	$0.14per talk minuteStart	42–100/100 documented fit 42% evidence coverage	Full research →

BEYOND A FEATURE CHECKMARK

How deep does the capability go?

Choose up to 3 for a closer look →

28 weighted checks: core capabilities contribute 32 points and workflow depth contributes 68. AI is included at the weight shown for this category. Expand Evidence for scope and sources, or ask for the same demonstration from each supplier.

Show checksShowing fewer rows does not change the score.

AI voice agents features for all 6 products
Feature / base weight	Retell AI 80–99/100 documented fit Core capabilities32 / 32 pts10/10 checks with established status Workflow depth47.9–66.8 / 68 pts12/18 checks with established status	ElevenLabs Agents 74–100/100 documented fit Core capabilities32 / 32 pts10/10 checks with established status Workflow depth41.6–68 / 68 pts9/18 checks with established status	Vapi 71–100/100 documented fit Core capabilities32 / 32 pts10/10 checks with established status Workflow depth38.6–68 / 68 pts8/18 checks with established status	Synthflow 66–100/100 documented fit Core capabilities25.6–32 / 32 pts8/10 checks with established status Workflow depth40.8–68 / 68 pts8/18 checks with established status	respond.io 47–100/100 documented fit Core capabilities28–32 / 32 pts9/10 checks with established status Workflow depth19.2–68 / 68 pts4/18 checks with established status	Bland AI 42–100/100 documented fit Core capabilities22.4–32 / 32 pts7/10 checks with established status Workflow depth20–68 / 68 pts4/18 checks with established status
Core capabilityTelephone deployment4 points · Deploy an agent to a documented phone or voice-calling channel.	Supported4 / 4 pts Evidence US numbers, SIP/web voice, tools, tests and knowledge. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Phone/web, tools, knowledge, transfer, evaluations and APIs. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Phone/web agents, tools, Squads, tests and business integrations. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Enterprise telephony, SIP, business integrations and knowledge sources. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence AI agents answer documented WhatsApp/VoIP voice calls; US number availability is not established by this feature. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Inbound phone deployment is documented. Official documentation · Oct 15, 2026Source
Core capabilityWeb voice or SIP connection4 points · Deploy voice through a browser interface or connect telephony using SIP.	Supported4 / 4 pts Evidence US numbers, SIP/web voice, tools, tests and knowledge. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Phone/web, tools, knowledge, transfer, evaluations and APIs. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Phone/web agents, tools, Squads, tests and business integrations. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Enterprise telephony, SIP, business integrations and knowledge sources. Official documentation · Oct 15, 2026Source	Not verified0–4 / 4 pts Evidence Not established in reviewed evidence; request this specific demonstration. Not verified · Oct 15, 2026Source	Not verified0–4 / 4 pts Evidence Not established in reviewed evidence; request this specific demonstration. Not verified · Oct 15, 2026Source
Depth checkConfigured live-call transfer5.6 points · Configure a production call transfer to a person or destination with observable transfer behavior. Ask for a demonstration Make a test phone call and transfer it to the configured human destination. A scenario to run with the supplier; not a completed product test.	Supported5.6 / 5.6 pts Evidence Transfer tools support human warm-transfer context. Official documentation · Oct 15, 2026Source	Supported5.6 / 5.6 pts Evidence The transfer tool routes calls to configured phone destinations. Official documentation · Oct 15, 2026Source	Supported5.6 / 5.6 pts Evidence Warm transfer provides prior conversation context and an operator summary. Official documentation · Oct 15, 2026Source	Supported5.6 / 5.6 pts Evidence Warm transfers support contextual introductions and fallback after unsuccessful transfer. Official documentation · Oct 15, 2026Source	Supported5.6 / 5.6 pts Evidence AI hands conversations or live voice calls to people with retained history. Official documentation · Oct 15, 2026Source	Supported5.6 / 5.6 pts Evidence Versioned pathways define transfer nodes, action variables and conditional boundaries. Official documentation · Oct 15, 2026Source
Depth checkFallback when call setup or agent fails6.4 points · Configure fallback handling when the voice agent or its telephony connection cannot serve a call. Ask for a demonstration Demonstrate the configured fallback for an unavailable agent or destination. A scenario to run with the supplier; not a completed product test.	Not verified0–6.4 / 6.4 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Not verified0–6.4 / 6.4 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Supported6.4 / 6.4 pts Evidence Configured fallback transcribers can take over after provider failure. Official documentation · Oct 15, 2026Source	Supported6.4 / 6.4 pts Evidence Warm transfers support contextual introductions and fallback after unsuccessful transfer. Official documentation · Oct 15, 2026Source	Not verified0–6.4 / 6.4 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Not verified0–6.4 / 6.4 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source
Core capabilityLive business actions4 points · Call a tool, webhook or API during a conversation.	Supported4 / 4 pts Evidence US numbers, SIP/web voice, tools, tests and knowledge. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Phone/web, tools, knowledge, transfer, evaluations and APIs. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Phone/web agents, tools, Squads, tests and business integrations. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Enterprise telephony, SIP, business integrations and knowledge sources. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Grounded AI, tools, handoff, testing and live monitoring. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Pathway nodes provide webhooks, transfers, knowledge and testing. Official documentation · Oct 15, 2026Source
Core capabilityHuman or specialist handoff4 points · Transfer a conversation to a human or another configured specialist.	Supported4 / 4 pts Evidence US numbers, SIP/web voice, tools, tests and knowledge. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Phone/web, tools, knowledge, transfer, evaluations and APIs. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Phone/web agents, tools, Squads, tests and business integrations. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Enterprise telephony, SIP, business integrations and knowledge sources. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Grounded AI, tools, handoff, testing and live monitoring. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Pathway nodes provide webhooks, transfers, knowledge and testing. Official documentation · Oct 15, 2026Source
Depth checkStructured action arguments5.6 points · Define tool parameters or structured fields so a business action receives inspectable values. Ask for a demonstration Call a booking or CRM tool and inspect the arguments sent for the requested action. A scenario to run with the supplier; not a completed product test.	Supported5.6 / 5.6 pts Evidence Custom tools define typed parameters and authentication headers separately. Official documentation · Oct 15, 2026Source	Supported5.6 / 5.6 pts Evidence Webhook tools define typed arguments and separate secrets/authentication. Official documentation · Oct 15, 2026Source	Supported5.6 / 5.6 pts Evidence Function tools define structured parameters for external actions. Official documentation · Oct 15, 2026Source	Supported5.6 / 5.6 pts Evidence Custom actions define variables, authenticated HTTP requests and result handling. Official documentation · Oct 15, 2026Source	Not verified0–5.6 / 5.6 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Supported5.6 / 5.6 pts Evidence Versioned pathways define transfer nodes, action variables and conditional boundaries. Official documentation · Oct 15, 2026Source
Depth checkContext-preserving handoff6.4 points · Pass conversation context or a summary to the next agent or human during a handoff. Ask for a demonstration Transfer a call after collecting information and show what the receiving person or agent receives. A scenario to run with the supplier; not a completed product test.	Supported6.4 / 6.4 pts Evidence Transfer tools support human warm-transfer context. Official documentation · Oct 15, 2026Source	Supported6.4 / 6.4 pts Evidence SIP transfers pass short context identifiers in headers; full summaries require another context transport. Official documentation · Oct 15, 2026Source	Supported6.4 / 6.4 pts Evidence Warm transfer provides prior conversation context and an operator summary. Official documentation · Oct 15, 2026Source	Supported6.4 / 6.4 pts Evidence Warm transfers support contextual introductions and fallback after unsuccessful transfer. Official documentation · Oct 15, 2026Source	Supported6.4 / 6.4 pts Evidence AI hands conversations or live voice calls to people with retained history. Official documentation · Oct 15, 2026Source	Not verified0–6.4 / 6.4 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source
Core capabilityPre-launch agent testing3.2 points · Test agent conversations before sending production traffic.	Supported3.2 / 3.2 pts Evidence US numbers, SIP/web voice, tools, tests and knowledge. Official documentation · Oct 15, 2026Source	Supported3.2 / 3.2 pts Evidence Phone/web, tools, knowledge, transfer, evaluations and APIs. Official documentation · Oct 15, 2026Source	Supported3.2 / 3.2 pts Evidence Phone/web agents, tools, Squads, tests and business integrations. Official documentation · Oct 15, 2026Source	Not verified0–3.2 / 3.2 pts Evidence Not established in reviewed evidence; request this specific demonstration. Not verified · Oct 15, 2026Source	Supported3.2 / 3.2 pts Evidence Grounded AI, tools, handoff, testing and live monitoring. Official documentation · Oct 15, 2026Source	Supported3.2 / 3.2 pts Evidence Pathway nodes provide webhooks, transfers, knowledge and testing. Official documentation · Oct 15, 2026Source
Core capabilityCall logs and monitoring3.2 points · Inspect call records, transcripts or operational analytics.	Supported3.2 / 3.2 pts Evidence US numbers, SIP/web voice, tools, tests and knowledge. Official documentation · Oct 15, 2026Source	Supported3.2 / 3.2 pts Evidence Phone/web, tools, knowledge, transfer, evaluations and APIs. Official documentation · Oct 15, 2026Source	Supported3.2 / 3.2 pts Evidence Phone/web agents, tools, Squads, tests and business integrations. Official documentation · Oct 15, 2026Source	Not verified0–3.2 / 3.2 pts Evidence Not established in reviewed evidence; request this specific demonstration. Not verified · Oct 15, 2026Source	Supported3.2 / 3.2 pts Evidence Grounded AI, tools, handoff, testing and live monitoring. Official documentation · Oct 15, 2026Source	Not verified0–3.2 / 3.2 pts Evidence Not established in reviewed evidence; request this specific demonstration. Not verified · Oct 15, 2026Source
Depth checkRepeatable scenario testing4.8 points · Run saved or generated test scenarios with outcomes or evaluation results. Ask for a demonstration Run the same successful and failing scenarios after a prompt change and inspect the results. A scenario to run with the supplier; not a completed product test.	Supported4.8 / 4.8 pts Evidence Simulation cases are rerunnable and graded against success criteria. Official documentation · Oct 15, 2026Source	Supported4.8 / 4.8 pts Evidence Saved simulation, next-reply and tool-call tests support repeat runs. Official documentation · Oct 15, 2026Source	Supported4.8 / 4.8 pts Evidence Simulation suites rerun scenarios with explicit pass/fail results. Official documentation · Oct 15, 2026Source	Supported4.8 / 4.8 pts Evidence Test Center saves scenario suites and evaluates outcomes; real action execution is excluded. Official documentation · Oct 15, 2026Source	Not verified0–4.8 / 4.8 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Not verified0–4.8 / 4.8 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source
Depth checkVersioned agent changes4.8 points · Inspect prior agent configurations and restore or select a previous version. Ask for a demonstration Change an agent instruction, then show how to return production to its earlier version. A scenario to run with the supplier; not a completed product test.	Supported4.8 / 4.8 pts Evidence Published versions and environment tags select the production configuration. Official documentation · Oct 15, 2026Source	Supported4.8 / 4.8 pts Evidence Versioned agent configuration supports selectable deployments. Official documentation · Oct 15, 2026Source	Supported4.8 / 4.8 pts Evidence Assistants retain published versions and restoration controls. Official documentation · Oct 15, 2026Source	Supported4.8 / 4.8 pts Evidence Agent configurations can be restored; action changes are outside version history. Official documentation · Oct 15, 2026Source	Not verified0–4.8 / 4.8 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Supported4.8 / 4.8 pts Evidence Versioned pathways define transfer nodes, action variables and conditional boundaries. Official documentation · Oct 15, 2026Source
Core capabilityDeveloper APIs or webhooks2.4 points · Manage agents or exchange conversation data using programmable interfaces.	Supported2.4 / 2.4 pts Evidence US numbers, SIP/web voice, tools, tests and knowledge. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Phone/web, tools, knowledge, transfer, evaluations and APIs. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Phone/web agents, tools, Squads, tests and business integrations. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Enterprise telephony, SIP, business integrations and knowledge sources. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Developer API, Zapier and Make on Growth. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Pathway nodes provide webhooks, transfers, knowledge and testing. Official documentation · Oct 15, 2026Source
Core capabilityBusiness or telephony connectors2.4 points · Use supplied CRM, calendar, contact-center or telephony integrations.	Supported2.4 / 2.4 pts Evidence US numbers, SIP/web voice, tools, tests and knowledge. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Phone/web, tools, knowledge, transfer, evaluations and APIs. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Phone/web agents, tools, Squads, tests and business integrations. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Enterprise telephony, SIP, business integrations and knowledge sources. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Developer API, Zapier and Make on Growth. Official documentation · Oct 15, 2026Source	Not verified0–2.4 / 2.4 pts Evidence Not established in reviewed evidence; request this specific demonstration. Not verified · Oct 15, 2026Source
Depth checkAuthenticated business-tool access3.2 points · Configure authentication or protected credentials for calling an external business tool. Ask for a demonstration Show a business tool connection with its credentials separated from the spoken conversation. A scenario to run with the supplier; not a completed product test.	Supported3.2 / 3.2 pts Evidence Custom tools define typed parameters and authentication headers separately. Official documentation · Oct 15, 2026Source	Supported3.2 / 3.2 pts Evidence Webhook tools define typed arguments and separate secrets/authentication. Official documentation · Oct 15, 2026Source	Not verified0–3.2 / 3.2 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Supported3.2 / 3.2 pts Evidence Custom actions define variables, authenticated HTTP requests and result handling. Official documentation · Oct 15, 2026Source	Not verified0–3.2 / 3.2 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Not verified0–3.2 / 3.2 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source
Depth checkStructured post-call event export4 points · Send or retrieve post-call structured results and call identifiers through an API or webhook. Ask for a demonstration Finish a test call and inspect its structured results in an external system. A scenario to run with the supplier; not a completed product test.	Supported4 / 4 pts Evidence Structured extraction results are accessible through call_analyzed events or Get Call. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Post-call webhooks expose analyzed conversation results. Official documentation · Oct 15, 2026Source	Not verified0–4 / 4 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Supported4 / 4 pts Evidence Post-call webhooks expose structured results, actions and call identifiers. Official documentation · Oct 15, 2026Source	Not verified0–4 / 4 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Not verified0–4 / 4 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source
Core capabilityBusiness knowledge grounding2.4 points · Provide business documents, website information or a knowledge source for answers.	Supported2.4 / 2.4 pts Evidence US numbers, SIP/web voice, tools, tests and knowledge. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Phone/web, tools, knowledge, transfer, evaluations and APIs. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Phone/web agents, tools, Squads, tests and business integrations. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Enterprise telephony, SIP, business integrations and knowledge sources. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Grounded AI, tools, handoff, testing and live monitoring. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Pathway nodes provide webhooks, transfers, knowledge and testing. Official documentation · Oct 15, 2026Source
Core capabilityConfigurable conversational behavior2.4 points · Configure agent instructions, dialogue paths or model behavior.	Supported2.4 / 2.4 pts Evidence US numbers, SIP/web voice, tools, tests and knowledge. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Phone/web, tools, knowledge, transfer, evaluations and APIs. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Phone/web agents, tools, Squads, tests and business integrations. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Configured conversational agents and fallback behavior. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Grounded AI, tools, handoff, testing and live monitoring. Official documentation · Oct 15, 2026Source	Supported2.4 / 2.4 pts Evidence Pathway nodes provide webhooks, transfers, knowledge and testing. Official documentation · Oct 15, 2026Source
Depth checkManage and refresh grounded knowledge3.2 points · Update or replace business knowledge used by the conversational agent and inspect its configured source. Ask for a demonstration Change a policy in a knowledge source and demonstrate how the agent’s source is refreshed. A scenario to run with the supplier; not a completed product test.	Supported3.2 / 3.2 pts Evidence Knowledge sources and indexing can be managed per agent. Official documentation · Oct 15, 2026Source	Supported3.2 / 3.2 pts Evidence Document management exposes source replacement and refresh. Official documentation · Oct 15, 2026Source	Not verified0–3.2 / 3.2 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Not verified0–3.2 / 3.2 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Supported3.2 / 3.2 pts Evidence Approved knowledge sources are synchronized; guardrails govern actions and escalation. Official documentation · Oct 15, 2026Source	Not verified0–3.2 / 3.2 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source
Depth checkExplicit action or escalation boundaries4 points · Configure a rule that restricts the agent’s permitted action or escalates an out-of-scope request. Ask for a demonstration Ask for a restricted action and show the configured refusal or human escalation. A scenario to run with the supplier; not a completed product test.	Supported4 / 4 pts Evidence Guardrails configure prohibited-content detection and replacement behavior. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Custom guardrails restrict unauthorized actions and sensitive answers. Official documentation · Oct 15, 2026Source	Not verified0–4 / 4 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Not verified0–4 / 4 pts Evidence This specific workflow has not been established by the reviewed evidence. Not verified · Oct 15, 2026Source	Supported4 / 4 pts Evidence Approved knowledge sources are synchronized; guardrails govern actions and escalation. Official documentation · Oct 15, 2026Source	Supported4 / 4 pts Evidence Versioned pathways define transfer nodes, action variables and conditional boundaries. Official documentation · Oct 15, 2026Source
Depth checkConfigure backup speech-synthesis providers2.5 points · Demonstrate "Configure backup speech-synthesis providers" with test data. Show the result, an exception case, and the required plan. Ask for a demonstration Demonstrate "Configure backup speech-synthesis providers" with test data. Show the result, an exception case, and the required plan. A scenario to run with the supplier; not a completed product test.	Supported2.5 / 2.5 pts Evidence Retell documents fallback to another TTS provider after provider failure. Product changelog documents feature; confirm current compatible voice models and local contract. Official documentation · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Supported2.5 / 2.5 pts Evidence An ordered backup-voice list can switch providers after a qualifying voice error. Not every outage/error is recoverable; provider, language and voice compatibility must be checked. Official documentation · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source
Depth checkConfigure backup speech-recognition providers2.5 points · Demonstrate "Configure backup speech-recognition providers" with test data. Show the result, an exception case, and the required plan. Ask for a demonstration Demonstrate "Configure backup speech-recognition providers" with test data. Show the result, an exception case, and the required plan. A scenario to run with the supplier; not a completed product test.	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Supported2.5 / 2.5 pts Evidence Manual ordered transcription fallbacks and optional automatic fallback are configurable. Fallback may route audio to another provider; supported languages and retention terms vary by provider. Official documentation · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source
Depth checkSet custom pronunciation for names and terms2.5 points · Set a custom pronunciation for a sample name. Compare two supported voice models and show any model or language limitations. Ask for a demonstration Set a custom pronunciation for a sample name. Compare two supported voice models and show any model or language limitations. A scenario to run with the supplier; not a completed product test.	Limited1.3 / 2.5 pts Evidence Custom pronunciation accepts word, phonetic alphabet and pronunciation values. Pronunciation support is voice/model-dependent; availability must be demonstrated for selected language/model. Official documentation · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source
Depth checkInspect latency by voice-pipeline component2.5 points · Demonstrate "Inspect latency by voice-pipeline component" with test data. Show the result, an exception case, and the required plan. Ask for a demonstration Demonstrate "Inspect latency by voice-pipeline component" with test data. Show the result, an exception case, and the required plan. A scenario to run with the supplier; not a completed product test.	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source
Depth checkConfigure reminders after caller silence2.5 points · Demonstrate "Configure reminders after caller silence" with test data. Show the result, an exception case, and the required plan. Ask for a demonstration Demonstrate "Configure reminders after caller silence" with test data. Show the result, an exception case, and the required plan. A scenario to run with the supplier; not a completed product test.	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source
Depth checkChoose a separate action when voicemail answers2.5 points · Demonstrate "Choose a separate action when voicemail answers" with test data. Show the result, an exception case, and the required plan. Ask for a demonstration Demonstrate "Choose a separate action when voicemail answers" with test data. Show the result, an exception case, and the required plan. A scenario to run with the supplier; not a completed product test.	Supported2.5 / 2.5 pts Evidence Voicemail handling can hang up, leave a recorded message or generate a personalized message. Product changelog; detection is not guaranteed and local outbound-calling permissions remain separate. Official documentation · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source
Depth checkRedact sensitive fields from call transcripts2.5 points · Demonstrate "Redact sensitive fields from call transcripts" with test data. Show the result, an exception case, and the required plan. Ask for a demonstration Demonstrate "Redact sensitive fields from call transcripts" with test data. Show the result, an exception case, and the required plan. A scenario to run with the supplier; not a completed product test.	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source
Depth checkVerify cryptographic post-call webhook signatures2.5 points · Demonstrate "Verify cryptographic post-call webhook signatures" with test data. Show the result, an exception case, and the required plan. Ask for a demonstration Demonstrate "Verify cryptographic post-call webhook signatures" with test data. Show the result, an exception case, and the required plan. A scenario to run with the supplier; not a completed product test.	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source	Not verified0–2.5 / 2.5 pts Evidence This specific check has not been verified. Missing evidence does not mean the capability is absent. Ask for a demonstration in the proposed plan. Not verified · Oct 21, 2026Source

WHAT WE MEASURE

The criteria behind the score.

The established 20 checks contribute 80 points. Eight new operational checks contribute 20 points. AI checks are included in their own group and are excluded when AI priority is set to zero. These assess documented capability scope, not measured performance. Adjust group priorities in the comparison tool.

See the complete calculation →

Phone and web voice deployment 20%
The required phone, SIP or browser deployment must be supported in the intended region.
Live actions and handoff 20%
An agent must perform the permitted task and transfer control without losing essential context.
Testing and call operations 16%
Testing, logs and version management make failures diagnosable and changes reviewable.
APIs and business connections 12%
Reliable business actions depend on usable APIs, connectors and authentication.
Configurable conversational AI 19.5%
Conversational behavior, knowledge use and model configuration should fit the task and its limits.
Operational detail 12.5%
Specific workflow, integration and control checks beyond the established rubric.

THE PRACTICAL BUYER’S GUIDE

How to choose ai voice agents.

AI voice agents can answer or place calls, interpret speech and interact with business systems. They are different from a phone system that transcribes a human conversation or summarizes voicemail. The practical question is whether an agent can complete your specific task with acceptable mistakes, delay and cost. A developer platform offers flexibility but leaves more implementation work with your team. A managed service or visual builder can reduce that work while introducing different packaging and operating constraints. Our comparison covers documented capabilities, not measured voice quality or claimed resolution rates. Before production, use the same test scenarios and acceptance rules for every shortlisted system, including calls where automation must stop.

Define one bounded task

Start with a task whose success can be observed in a business system: collect a callback request, route an inquiry or schedule an eligible appointment. Write down what the agent may change, which facts it must confirm and when a person must take over. Avoid combining sales, account servicing and exception handling into the first release. Provide approved information and explicit examples of unavailable answers. Evaluate the agent on task completion, not merely fluent conversation. A convincing voice can still record the wrong date or promise an unavailable service. Require the system to distinguish a successful action from a request that was only submitted or queued.

Evaluate failure paths as carefully as the demo

Give each vendor the same scenarios, including background noise, an interruption, an unusual name, an ambiguous time, an API timeout and a caller who changes their mind. Check what happens when a transfer target is busy or closed. Repeat tests after changing a prompt, model, voice or tool definition; any of these can alter the experience. Inspect call logs and the resulting business record together. A transcript alone does not establish whether an action completed correctly. Decide in advance which failures block launch, which require a narrower use case and which can be monitored during a limited rollout with staff available.

Build the complete per-task cost

A published per-minute rate may cover only the platform. Depending on the provider, voice generation, language models, telephony, transfers, recordings or concurrency can add charges. Calculate low, expected and busy-period scenarios using realistic call duration and the share of calls needing a person. Include implementation and ongoing review time. Compare cost per correctly completed task as well as cost per minute; a short failed call is not a bargain. Sales-led plans may have annual commitments or volume thresholds that make them unsuitable for a small pilot. Ask for written treatment of failed calls, silence, retries and transferred call legs before signing.

What changes the total cost?

Platform minutes plus any separate model, voice and telephony charges.
Concurrent calls, transferred call legs, recordings and monitoring features.
Implementation, evaluation, human escalation staffing and minimum commitments.

Where purchases go wrong

Equating a natural voice demo with verified task completion.
Comparing an orchestration-only price with an all-inclusive minute price.
Launching without a staffed escalation route and an action audit trail.

Bring these questions to the demo

Show the external business record after the agent completes our task.
How does the agent respond when a tool times out or returns conflicting data?
What happens when a person is requested but the transfer destination cannot answer?
Which parts of our example call appear as separate billable line items?

Questions buyers ask

Is an AI receptionist a full business phone system?

Usually it is a role within a larger calling setup. Check number ownership, routing, employee calling and administration separately. A receptionist can complement a phone system; its availability does not establish that it replaces all PBX capabilities.

Can we compare agents by latency or accuracy here?

The directory does not claim hands-on benchmark results. Published capabilities indicate what can be configured, not how reliably it performs for your callers. Use a controlled pilot with your audio conditions, knowledge, integrations and outcome definitions before making performance claims.

Why do quoted minute prices differ so much?

Providers include different components and impose different commitments. Some charge platform orchestration separately from models and carriers; others bundle more of the stack. Normalize a complete call scenario and ask for a sample invoice, including transfers and expected peak concurrency.

AI voice agents

Compare the options.

Retell AI

ElevenLabs Agents

Vapi

Synthflow

respond.io

Bland AI