Month: October 2026

TL;DR: Several AI voice platforms can write to a CRM while the caller is still on the line, and they do it in very different ways. Developer-configurable platforms such as Retell AI, Vapi and Bland AI hand you a function call and let you build the write. Packaged agent builders such as Synthflow AI and Thoughtly ship the integration, and their native CRM paths differ: Thoughtly documents mid-call CRM actions, while Synthflow's native CRM integration writes at hang-up, and Synthflow can still write mid-call through custom or MCP actions. Business phone platforms such as Aircall, Dialpad and CloudTalk bolt an agent onto an existing phone system, and one of them writes to the CRM only after the call ends. Enterprise suites such as PolyAI, Cognigy and Talkdesk treat the write as one step in a governed flow. Salesforce Agentforce writes natively, inside Salesforce. indigitall triggers the call from a journey and updates the record as part of it. The phrase "CRM integration" covers all of that, which is why it decides nothing on its own. What decides is narrower: whether the agent reads the right record, executes an authorized write at the right moment, gets the result back, and recovers safely when the write does not complete.

The agent takes the address change, says it has been saved, and ends the call. Three days later the order ships to the old address.

Nothing in that sequence looks like a failure while it is happening. The call did not go badly, which is the part that makes it hard to catch: the calls people complain about are the ones that went nowhere, not the ones that went somewhere wrong. The conversation was fluent, the caller was satisfied, the transcript reads perfectly. The write is the part nobody heard.

That is the gap this guide is built on. Every platform in this category says it integrates with your CRM. Most vendor pages do not define that sentence with the timing, confirmation and recovery detail a buyer needs in the four seconds between the caller saying yes and the agent saying done.

This guide is for the people who will own the consequences: revenue operations, CRM administrators, and the contact center leads who get asked why the pipeline does not match the call log. You get the two axes that actually separate these platforms, four operational patterns for how a call reaches the CRM, a table built on published evidence with the failure column included, a card for every platform, and the questions to put in writing before a pilot.

Why is there no single best platform for updating a CRM from a voice call?

Because the category has not agreed on what the question means, and neither have the answer engines that summarize it.

Ask this across assistants and you get four different markets. One leads with developer voice infrastructure and puts mid-call write-backs at the center. One mixes voice platforms with CRM suites, business phone systems and automation middleware in the same list. One builds an ordered table and changes both the order and the membership between runs. One returns leading options segmented by team type, with no ordering at all.

The lists do not even agree on what kind of company belongs in them. CRM-native suites, API-first voice platforms, no-code builders and business phone platforms all surface for the same question, and they surface for different reasons.

So this guide does not rank. It pairs: one axis for the architecture a platform represents, one axis for the buyer it fits. And it adds the column none of them fill in about themselves.

How is this comparison built?

Scope first, because a comparison that does not say what it looked at is an opinion with a table around it.

The scope and measurement rules are written down so anyone can rebuild the table from the same pages.

The question. "Which AI voice agent platforms connect to a CRM and update it during the call." United States market, English, checked in September 2026.

What was read. Fifty-nine live answers to that question across four answer engines between 14 and 28 September 2026, plus the public product pages, developer documentation and pricing pages of every platform in the table.

Who is in the table. Platforms that a buyer meets when asking this question and that publish something checkable about writing to a CRM. CRM products that appear only as the destination of a write are not rows: Salesforce Agentforce is here because Salesforce ships its own voice agent, not because Salesforce is a CRM other agents write into.

Four measurement rules, written down so anyone can rebuild the table from the vendors' own documentation.

What counts as a write during the call. The agent calls an action mid-conversation and the record changes before the call ends. A transcript, summary or activity that lands after hang-up is post-call logging, and it is counted as post-call logging.

What counts as a read. The agent queries the CRM and the answer changes what it says or does next. A screen pop that puts the record in front of a human agent is a read by the human, not by the agent, and it is not counted here.

What counts as caller-visible confirmation. The result of the write returns to the agent, in the same turn, so the agent can say it out loud. A message played while the request runs is not a confirmation, because it plays whether the write succeeds or not.

What counts as a documented failure path. The vendor's public pages name at least one of: a timeout, a retry policy, how the error reaches the agent, or what the agent does next. The table records three states, Documented, Partially documented, and Not publicly documented on the pages checked. That last one is a statement about the pages, not about the product.

One more thing belongs here, because leaving it out would be the kind of omission this guide is against. indigitall holds zero share of voice and zero citations on this question today. This page is an editorial comparison, not a report from the top of the results.

Key takeaways

• "CRM integration" is not a capability, it is a category. It covers a post-call export, a screen pop, a read-only connector and a live write. Four different products answer yes to the same question, so the yes decides nothing.
• The write has a moment, and the moment is the whole thing. Ask whether the record changes before the caller hangs up or after, because one of those removes the follow-up and the other schedules it.
• A read that only reaches a person is not a read. If the record appears on a human agent's screen, the voice agent did not use it, and nothing it says next is informed by it.
• Confirmation out loud has to come after the result comes back. An agent that says "saved" before the write returns is not lying on purpose, it is just speaking in the wrong order, and nobody hears the difference.
• The failure path is the column nobody publishes. A vendor that documents a timeout, a retry or how the error reaches the agent has thought about the bad call. Most public pages in this category document only the good one.

Is CRM integration a yes-or-no feature?

No, and treating it as one is what produces the address that never changed. There are two independent axes, and a platform can sit high on one and low on the other.

The read axis, from least to most capable:

• The agent does not read the CRM at all.
• The agent receives context assembled before the call, usually by whatever launched it.
• The agent queries the CRM during the conversation.
• The agent queries, gets the result back, and uses it to decide the next step.

The write axis, same direction:

• No structured write at all.
• A transcript, summary or activity is stored after hang-up.
• A task or queue item is created and a person makes the CRM change.
• A webhook, workflow or batch process updates the CRM after the call.
• A function writes during the call and returns the result to the agent.

The bottom of the read axis and the top of the write axis can coexist, and that combination is the dangerous one: an agent that writes confidently into a record it never checked is how a duplicate contact gets created instead of an existing one being updated. Vapi's own documentation is unusually blunt about the ordering this requires, telling builders to "always use 'getGHLContact' before 'createGHLContact'".

🖼️ Visual suggestion, diagram: the two axes as a grid, read on one side and write on the other, with the five write levels and four read levels labeled and the platform groups placed on it. INTERNO: inserta aqui diagram.
INTERNO, brief de diseno: imagen nueva. Una cuadricula simple sobre blanco, sin iconos, con los ejes rotulados y los seis grupos de arquitectura colocados como etiquetas, no como puntos. Azul de casa para los rotulos de eje y gris para la rejilla. Lo que tiene que quedar claro al ojear es que son dos ejes y no una escala.
Alt: A grid showing the CRM read axis against the CRM write axis, with the six architecture groups placed according to what their documentation describes.
Pie: A platform can sit at the top of the write axis and the bottom of the read axis, and that combination is how an agent creates a second contact instead of updating the first.

What are the four ways a call ends up in the CRM?

Four operational patterns. Most platforms support more than one, and naming which one is running in your workflow is the single most useful thing a demo can establish.

🖼️ Visual suggestion, diagram: the four patterns laid on one call timeline, from the caller saying yes to the record changing, so the distance between the two is visible in each case. INTERNO: inserta aqui diagram.
INTERNO, brief de diseno: imagen nueva. Cuatro carriles horizontales sobre blanco, todos con la misma linea de tiempo debajo. En cada carril, el punto en que el registro cambia. El carril de la escritura en llamada cierra dentro de la llamada, los otros tres despues. Sin iconos.
Alt: The four ways a call reaches the CRM, laid on one timeline, showing where in the call each one changes the record.
Pie: Only one of the four closes inside the call. The other three move the work rather than removing it.

Post-call logging. The transcript, the summary or the call activity is stored against the record after hang-up. It is the oldest pattern and the most common, and it is the one most often described as CRM integration. CloudTalk states it plainly: "When a VoiceAgent handles a call, it processes the transcript after the call ends… CloudTalk then sends this JSON result to your connected CRM as part of a call log."

Human-mediated update. The system creates a task or a queue item and a person completes the CRM change. Strictly this is not a write pattern, it is a destination, and it is the honest answer for anything that needs judgment. It moves the work. It does not remove it. Count it as cost.

Deferred machine write. A webhook, workflow or batch process updates the CRM after the call ends, with no person involved. Faster than the previous one, and invisible when it fails, which is why retry and deduplication behavior matters more here than anywhere else. Retell AI publishes both for its webhooks: a ten-second timeout with up to three retries, and a recommendation to deduplicate on the combination of event and call identifier.

Synchronous live write. The agent calls a CRM action during the conversation, receives a result, and can confirm it to the caller before hanging up. This is what the question is actually asking about, and it is the only pattern where the caller can be told the truth about what happened.

The patterns stack, and the shift behind all four is the same one described in the guide to agentic AI: the agent stopped being a thing that answers and became a thing that acts. A platform can write live for the fields it has actions for, defer the summary to a webhook, and drop a task for anything the agent could not resolve. That is a good design. What is not good is not knowing which one is carrying which field.

How do the engines group this category today?

Differently, and that is the finding that shaped this page.

The first engine leads with architecture descriptors and a best-for label per platform, putting developer voice infrastructure at the front. Another alternates between ordered tables and selection by CRM, and changes the set between runs. A third segments its leading options by team type and puts business phone platforms ahead of the CRM suites. The fourth mixes platforms, suites and explanatory sources without assigning a fit label at all.

The labels they do produce are specific enough to be worth reusing, and this guide reuses them and invents none of its own. Retell AI arrives as "developer-friendly voice infrastructure supporting sub-second response times and custom mid-call CRM write-backs via function calling and APIs". Synthflow AI arrives as an "out-of-the-box platform that handles live appointment booking, lead qualification, and direct CRM updates mid-conversation without complex coding". Thoughtly arrives as "designed for operational teams to build voice agents that interact with databases and execute live CRM field updates while the caller is still on the line".

There is no single market-wide ordering to reproduce, so this comparison pairs platforms by architecture and buyer fit instead of inventing a universal ranking.

The comparison table

Read the last two columns first. The failure column is the one no vendor answers about itself, and the verify column is the one that decides the pilot.

One row has no engine label, and that is deliberate. indigitall does not appear in the live answers to this question, so there is no wording from an engine to put in that cell. Writing one would falsify the whole column, so the cell says what it is.

Where each cell comes from is named, one card at a time. Every platform card below ends with the pages this comparison was built from, so the failure column can be checked rather than trusted. The pages are named and not linked: several of these vendors compete with the publisher of this guide, and naming a page a reader can find is enough to audit a claim.

And the failure column is a statement about published pages, not about products. "Not publicly documented on the pages checked" means the vendor has not written it down where a buyer can read it. Several of these platforms almost certainly handle timeouts and retries well and have simply never published how. The only way to find out is to ask, with the question already written, which is what the last column is for.

The platforms, one by one

Thirteen platforms, grouped by architecture and alphabetical inside each group. The wider category they all sit in is mapped in the guide to AI agent software. The group is not a quality tier: it is a description of who builds the write path, which is the decision underneath all the others.

Bland AI

A programmable voice platform with built-in tools and custom API actions for live calls, sold to teams running outbound at scale.

🖼️ Visual suggestion, screenshot [external]: the Bland AI page at https://www.bland.ai/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://www.bland.ai/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: Bland AI home page headlined on voice AI for regulated industries, with the latency figure and the compliance badges below.
Pie: It sells infrastructure and control, not a finished workflow, which is why the CRM write is yours to define. Bland AI, captured in 2026.

Best fit: enterprise-scale outbound calling where the call itself is the workflow and the CRM write is one of its steps.

What it does well: it reads the record while the conversation is happening. Its documentation describes custom queries that "read Salesforce data over the same connection, so the agent has the record open while the caller is still on the line", and tool results can branch the call flow.

Key features:

• Custom tools against your own endpoint, with a defined input schema
• Prebuilt actions to create contacts, leads and opportunities
• Branching on whether the tool succeeded, with comparison operators
• A spoken holding phrase while a tool runs

Watch out for: there are two execution modes and only one of them reports back. In fire-and-forget the request goes out and the call continues, so the agent never learns whether the write landed. Its own documentation also warns that variables referenced too early "might have null values".

Pricing: published. A stated 0.14 US dollars per minute on the entry tier and 0.12 plus a 299 US dollar monthly fee on the next, with transfers priced separately and enterprise custom. Rates as published in 2026.

Doesn't fit if: you want the CRM logic delivered in the box, because here the logic is yours.

Public evidence: Bland tools API reference and custom tools tutorial (write path and timeout), the Salesforce integration page (objects), the pricing page (rates). Checked September 2026.

Retell AI

Voice infrastructure sold to developers, with native connectors layered on top of a function-calling model.

🖼️ Visual suggestion, screenshot [external]: the Retell AI page at https://www.retellai.com/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://www.retellai.com/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: Retell AI home page headlined as the number one AI voice agent platform for automating phone calls, with the drag and drop canvas shown.
Pie: The connectors sit on top of a function-calling model, so what the agent can write is whatever you wire underneath. Retell AI, captured in 2026.

Best fit: teams that want mid-call write-backs and are comfortable owning the parts the connectors do not cover.

What it does well: it publishes unusually concrete detail about what happens when a custom function call goes wrong, which makes the economics and the risk easy to model before the demo, which is when it is useful.

Key features:

• Tools the agent can call before, during or after a call
• Native connectors for five named CRMs, plus helpdesk and calendar
• Contact property updates, notes and tasks created mid-conversation
• The agent continues speaking after a function returns, so it can read the result back

Watch out for: the native tools have their own timeout, stated as 3 to 14 seconds depending on the provider, which is a much shorter budget than the custom-function timeout you control. A slow CRM will hit the connector limit first.

Pricing: published, at a stated 0.07 to 0.31 US dollars per minute depending on models and add-ons. Rates as published in 2026.

Doesn't fit if: nobody on the team owns the integration, because the flexibility that makes it cheap is the flexibility that leaves it unfinished on day one.

Public evidence: Retell custom function reference (timeout, retries, error handling), the integrations overview and the HubSpot integration page (mid-call writes), the webhook overview (post-call contract), the pricing page. Checked September 2026.

Vapi

A developer-first voice platform where the CRM layer is something you build, not something you buy.

🖼️ Visual suggestion, screenshot [external]: the Vapi page at https://vapi.ai/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://vapi.ai/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: Vapi home page headlined "Voice agents for builders", with the developer entry points listed.
Pie: The page addresses builders, not buyers, which is an accurate description of where the CRM layer comes from here. Vapi, captured in 2026.

Best fit: technical teams with an existing backend that already knows how to write to their CRM.

What it does well: it is explicit about the ordering a safe write requires, instructing builders to look up a contact before creating one and to check availability before creating an event. That is the duplicate-contact problem, solved in the documentation, where a builder finds it before the first duplicate.

Key features:

• Function tools that call your server, with a synchronous option that waits for the response
• Four configurable spoken messages, including one for a failed request
• Result matching by tool call identifier, so the agent knows which answer belongs to which request
• MCP tools and client-side tools alongside the server route

Watch out for: the error contract is strict and counterintuitive. The documentation states that you must "always respond with HTTP 200, even when reporting a tool failure", because "any other status code is ignored completely, and the assistant reports 'no result returned' instead of speaking your error message". Propagate your CRM's 500 and the caller hears nothing useful.

Pricing: published. A stated 0.05 US dollars per minute of platform hosting, with support packages from 29 US dollars a month and provider costs listed separately. Rates as published in 2026.

Doesn't fit if: you expected a CRM connector in the box, because for the two largest CRMs the documented route is your own endpoint.

Public evidence: Vapi tools and tool-calling documentation (result handling and the HTTP 200 error contract), the API request tool page (timeout and backoff), the GoHighLevel tool page (objects and ordering), the pricing page. Checked September 2026.

Synthflow AI

A packaged agent builder sold on getting a working voice agent live without engineering, with a large catalog of named integrations behind it.

🖼️ Visual suggestion, screenshot [external]: the Synthflow AI page at https://synthflow.ai/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://synthflow.ai/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: Synthflow AI home page headlined on enterprise-ready voice agents for automated phone calls, with the integration catalog below.
Pie: The promise is a working agent without engineering, and the catalog is the proof it offers. Synthflow AI, captured in 2026.

Best fit: agencies and operations teams that want the agent configured, not coded.

What it does well: the integration catalog is one of the largest here, at a stated 200 or more tools, and the custom action model is genuinely usable during a call, with returned fields becoming variables the rest of the conversation can reference.

Key features:

• HTTP custom actions that run before or during a call
• MCP actions that let the model invoke an external server's tools
• Real-time calendar checking and booking during the call
• Post-call SMS, email and information extraction

Watch out for: the native CRM integrations are post-call. Its own pages describe contact details, case updates and custom fields updating "the moment a call ends", which is hang-up, not mid-call. Anything that has to land while the caller is listening goes through a custom action you configure and maintain.

Pricing: only the enterprise floor is published, stated as contracts starting at 30,000 US dollars annually, scoped on volume, concurrency, telephony, integrations and security. No per-minute rate is published. As published in 2026.

Doesn't fit if: you need to compare cost per completed outcome before a sales conversation, because there is no published rate to model with.

Public evidence: Synthflow custom actions documentation (mid-call actions and failure messages), the actions overview, the HubSpot page and the Salesforce integration page (native post-call writes), the pricing page. Checked September 2026.

Thoughtly

A voice agent platform built around a node-based flow, with mid-call actions as a first-class feature and not an escape hatch.

🖼️ Visual suggestion, screenshot [external]: the Thoughtly page at https://thoughtly.com/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://thoughtly.com/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: Thoughtly home page headlined "Voice AI for the leads you can't miss.", with the node-based flow shown.
Pie: The flow is visible on the home page, which matters here: the mid-call action is a node you can see, not a hidden step. Thoughtly, captured in 2026.

Best fit: revenue and operations teams that want the write, the branch and the fallback designed in the same place.

What it does well: it documents the order of operations inside a single node, so a builder can see exactly when the action runs, when variables update and when the outcome is evaluated. It also publishes a small safeguard that matters more than it sounds: when updating records it "avoids sending empty fields where supported", which is how a partial answer stops wiping a populated field.

Key features:

• Mid-call actions and post-call automations across eight named CRMs
• Create, retrieve, update, delete and search on contacts, plus deal retrieval and update
• Inbound CRM sync that pulls records and their attributes into the platform
• Documented branching on action status, with timeout and fallback paths

Watch out for: the retry is a branch you design, not a platform behavior. Its own list of common mistakes includes "forgetting to cap retries or offer a human hand-off", which tells you who owns that decision.

Pricing: published. A stated 500 US dollars a month entry plan, with higher tiers quoted, and usage above the monthly allowance billed in arrears. Rates as published in 2026.

Doesn't fit if: you want a platform that decides the recovery behavior for you, because here it is part of the build.

Public evidence: Thoughtly actions documentation (mid-call order of operations, status branching, retry and fallback), the HubSpot and Salesforce CRM integration pages, the CRM sync page, the pricing page. Checked September 2026.

Aircall

A business phone system with an AI voice agent sold as a separate product on top of it.

🖼️ Visual suggestion, screenshot [external]: the Aircall page at https://aircall.io/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://aircall.io/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: Aircall home page leading with customer communications and an AI that handles interactions end to end, with the phone system beneath it.
Pie: The phone system is the product and the agent is sold on top of it, which is why the published action list is the agent's, not the platform's. Aircall, captured in 2026.

Best fit: growing teams whose phone system is already the center of the operation and who want the agent inside it.

What it does well: it runs the lookup early. An action can be set to "run automatically before the caller even finishes hearing the greeting", matching on the phone number, and the documentation states that if no record matches "the call continues normally" instead of stalling.

Key features:

• AI Actions that execute during live calls
• Named actions across four connectors, including create and update on contacts and companies
• Configurable confirmation before an action executes
• Case creation, update and resolution on the service side

Watch out for: the headline integration count belongs to the phone system, not to the agent. The agent's published action list covers four connectors plus native messaging, so the question for the demo is whether your specific write is on that list. The documentation is also candid that the messaging action "doesn't report back to the agent whether the message sent successfully", which is a useful clue about what confirmation means here.

Pricing: published. Phone licenses from a stated 30 US dollars per license per month with a three-license minimum, and the voice agent priced separately from a stated 100 US dollars monthly minimum, with per-minute rates above the included allowance. Rates as published in 2026.

Doesn't fit if: the write you need is not one of the agent's published actions, because the rest of the catalog is not reachable from inside the call.

Public evidence: The Aircall AI Actions support article (the published action list, caller context and the messaging caveat), the AI voice agent product page, the pricing page. Checked September 2026.

CloudTalk

A business phone platform whose AI agent handles the conversation and hands the outcome to the CRM afterwards.

🖼️ Visual suggestion, screenshot [external]: the CloudTalk page at https://www.cloudtalk.io/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://www.cloudtalk.io/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: CloudTalk home page headlined "Every call answered, 24/7. By your team, or by AI.", with the call handling blocks below.
Pie: Answering is the promise on the page. What happens to the record afterwards is a separate claim, made elsewhere. CloudTalk, captured in 2026.

Best fit: teams that want call handling automated and the CRM kept tidy, not the CRM changed mid-conversation.

What it does well: it is unusually specific about how CRM logging goes wrong, publishing a list that includes duplicate contacts when matching is loose, appointment times arriving as text where a structured date was expected, and a disposition mapping skipped per campaign so the field is silently lost. That is more operational honesty than the category usually offers.

Key features:

• Post-call export of the agent's structured result to seven named CRMs
• A configurable extraction step that decides which fields are captured
• Tools that make API requests to third-party systems during a call
• A spoken holding message while a tool runs

Watch out for: the native CRM path is explicitly after the call, and the in-call path is a generic API tool you build. Two different things share the phrase CRM integration on the same product.

Pricing: published. Phone plans from a stated 19 euros per user per month billed annually with a three-user minimum, and the AI agent tiers priced separately with included minutes. Rates as published in 2026.

Doesn't fit if: the caller has to be told the record changed before hanging up, because the native path cannot do that.

Public evidence: The CloudTalk article on exporting VoiceAgent call results to CRM integrations (post-call export), the Tools for VoiceAgent article (in-call API calls), the company's own article on CRM logging failure modes, the pricing page. Checked September 2026.

Dialpad

A phone and contact center platform with an agentic layer that executes units of work against connected systems.

🖼️ Visual suggestion, screenshot [external]: the Dialpad page at https://www.dialpad.com/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://www.dialpad.com/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: Dialpad home page headlined "First in AI. Best in Agentic.", with the contact center product blocks below it.
Pie: The word on the page is agentic, which here means units of work executed against connected systems rather than a conversation alone. Dialpad, captured in 2026.

Best fit: sales teams that want conversation intelligence and an agent working inside the same stack.

What it does well: it names its actions as discrete operations, not as an integration in general, and it publishes a governance layer with rules a buyer can recognize, including instructions to ask for confirmation before acting and to escalate in defined situations instead of acting.

Key features:

• Data operations such as create contact, update lead and log call note
• Ten named CRM and calendar connectors for the agentic layer
• Guardrails that define when the agent must confirm and when it must escalate
• Automatic call logging to the matching record, separate from the agentic writes

Watch out for: two mechanisms coexist and they are easy to confuse. The agentic data operations run in real time. The classic call logging happens after the call. Checking which one carries your field is the whole exercise.

Pricing: not published as rates. The pricing page describes a credit model in which a conversation is billable only when the agent retrieves information or executes a real action, without publishing the price of a credit.

Doesn't fit if: you need to model cost per completed outcome before committing, because the unit price is not public.

Public evidence: The Dialpad help article on creating an agentic AI agent (data operations, connectors and guardrails), the contact center integrations page, the pricing page. Checked September 2026.

Cognigy

An enterprise conversational AI platform where the voice call is a flow, and the CRM write is a node inside it.

🖼️ Visual suggestion, screenshot [external]: the Cognigy page at https://www.cognigy.com/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://www.cognigy.com/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: Cognigy home page headlined "Say Goodbye to Poor Customer Service", with the enterprise conversational AI positioning below.
Pie: It is sold to contact center owners, not to developers, which is why the write lives inside a governed flow. Cognigy, captured in 2026.

Best fit: enterprise and API-led teams standardising voice and digital on one orchestration layer.

What it does well: it is one of the more explicit platforms here about failure handling, including how the tool result returns to the agent. Its own documentation describes sending an answer back so the agent can "inform about the success, failure, or other details of the tool", which closes the loop between the write and what the caller hears.

Key features:

• Tools on the agent node, plus a generic HTTP request node, both executing during the flow
• A named extension for the largest CRM, with create and retrieve on cases
• Required fields surfaced in the builder, with picklists pulled from the connected instance
• Call failover that can transfer the call on a runtime error

Watch out for: the timeout depends on where you are hosted, stated as 15 seconds on shared and 8 seconds on dedicated and on-premises, with a warning that raising it further "may cause delays and affect AI Agents' performance". A slow CRM and the wrong environment is a bad combination discovered late.

Pricing: not published. There is no public pricing page.

Doesn't fit if: the deployment is small enough that a full orchestration layer is more process than the problem justifies.

Public evidence: The Cognigy HTTP request node reference and the support article on handling HTTP request errors (timeout, retries, error branching), the AI agent node reference, the resolve tool action reference (returning the result to the agent), the Voice Gateway endpoint reference (call failover). Checked September 2026.

PolyAI

An enterprise voice platform where agent behavior is written in code and the CRM call is a function.

🖼️ Visual suggestion, screenshot [external]: the PolyAI page at https://poly.ai/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://poly.ai/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: PolyAI home page headlined "Own every conversation", with the enterprise dialog agent positioning below it.
Pie: Agent behavior written in code is the differentiator it leads with, and it is also what makes the CRM call a function. PolyAI, captured in 2026.

Best fit: large contact center operations where the volume justifies an integration project.

What it does well: it publishes a worked example of the write itself, including the endpoint, the object and the fields, which is rarer than it should be and makes the effort estimable.

Key features:

• Functions in code that call declared HTTP endpoints, with per-environment base URLs and authentication
• Case creation and record management during calls
• Named integrations across four CRMs and a long list of vertical systems
• Documented handoff to live agents

Watch out for: what is published about failure describes operations, not runtime. It sets thresholds for when to escalate to its own support, including repeated errors that persist after retrying and an error rate above 5 percent across conversations, but the timeout and retry behavior of a single write is not described.

Pricing: not published. The page states only that ongoing use is priced per minute.

Doesn't fit if: you need the write configured for you, because the documented route is a function someone writes and maintains.

Public evidence: The PolyAI integrations introduction and Salesforce integration page (the worked write, with endpoint, object and fields), the common issues page (operational thresholds), the pricing page. Checked September 2026.

Talkdesk

A contact center platform with an agent builder layered on top, sold around skills the agent can invoke.

🖼️ Visual suggestion, screenshot [external]: the Talkdesk page at https://www.talkdesk.com/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://www.talkdesk.com/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: Talkdesk home page headlined "The hybrid workforce starts here.", with the contact center platform blocks below.
Pie: The frame is a workforce of people and agents together, which is why its unit is a skill the agent can invoke. Talkdesk, captured in 2026.

Best fit: contact center organizations standardising on one vendor for routing, agents and reporting.

What it does well: it is explicit about the limits of its own structure, publishing the maximum depth of the agent hierarchy, the number of agents per level and the number of skills per agent. Knowing the ceiling before designing against it saves a rebuild.

Key features:

• Integration skills that retrieve, update and send data to third-party systems
• Six named CRM connectors, with two distinct paths into the largest one
• A preview skill type for connecting external tool servers
• Structured escalation that passes context to a human agent

Watch out for: the confirmation does not reach the caller by default. The documentation states that a skill's output "is only accessible to the Action Agent that invoked it", and that the agent's instructions must say how to explain that output. Left unwritten, the write happens and the caller hears nothing about it.

Pricing: published, by seat. Stated at 85, 105, 165 and 225 US dollars per user per month across four tiers, with the agent capabilities presented inside the plans and not priced per minute. Rates as published in 2026.

Doesn't fit if: you want per-minute economics for an automated workflow, because the published model is per seat.

Public evidence: The Talkdesk AI Agent Platform overview, the integration skill page and the core concepts page (skill output and structural limits), the pricing page. The support pages returned a 403 to a logged-out check on 29 September 2026. Checked September 2026.

Salesforce Agentforce

A CRM vendor's own agent, with voice added, writing inside the system of record, with no connector in between.

🖼️ Visual suggestion, screenshot [external]: the Salesforce Agentforce page at https://www.salesforce.com/agentforce/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://www.salesforce.com/agentforce/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: Salesforce Agentforce product page describing an AI agent platform delivering autonomous support at enterprise scale.
Pie: There is no connector in this story because there is nothing to connect to: the record and the agent are the same product. Salesforce Agentforce, captured in 2026.

Best fit: organizations where Salesforce is already the system of record and the goal is fewer moving parts.

What it does well: there is no integration layer to build, break or maintain for the primary case. Its own description is that voice agents can update records, create cases and trigger workflows "all in real time and within Salesforce", and the escalation path hands the human "the full transcript and a prepopulated case already attached".

Key features:

• Standard actions including create case, close case, update record, add case comment and log a call
• Custom actions built with flows or code
• External service actions for reaching systems outside Salesforce
• A published rate card for voice minutes and voice actions

Watch out for: native means native. A write into a different CRM goes through an external service action someone builds against an interface description, which puts you back in the same position as the developer-configurable platforms with a different bill.

Pricing: published, through a credit model. A stated 500 US dollars per 100,000 credits and 2 US dollars per conversation, with published multipliers for voice minutes and voice actions, and credits that do not roll over. Rates as published in 2026.

Doesn't fit if: your system of record is anything else, because the advantage that makes it strong disappears the moment the write leaves the platform.

Public evidence: The Agentforce Voice product page (real-time actions), the standard agent actions reference (object coverage), the developer post on Apex action best practice (the error-handling guidance), the pricing page and the published rate cards. Named and not linked, because the vendor is on this client's competitor list. Checked September 2026.

indigitall

Voice as one step inside a customer journey, not a standalone agent, with the record update happening as part of that journey.

🖼️ Visual suggestion, screenshot [external]: the indigitall page at https://indigitall.com/en/ai-voice/, framed on the top third. INTERNO: inserta aqui screenshot.
INTERNO, brief de diseno: captura externa de https://indigitall.com/en/ai-voice/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, sustituye en el pie el ano por el dia de la captura.
Alt: indigitall AI Voice page showing the outbound call as a step inside a customer journey, with the trigger and the actions listed.
Pie: The call is drawn with a trigger before it and a record update after it, which is the shape this section describes. indigitall, captured in 2026.

Best fit: organizations already running reminders and follow-ups over several channels that want the call to share the same customer profile and the same rules.

What it does well: the call does not start with the call. A journey step or a prediction triggers it, the agent acts on the connected systems, and the follow-up message goes out from the same place, so the record and the messaging stay in one profile instead of two.

Key features:

• Outbound calls triggered by a journey step or a prediction
• Actions on connected systems through MCP, described as fetching data, calling functions and writing results back
• Warm transfer carrying the transcript, the detected intent and the steps already taken
• The same customer profile across push, email, SMS, WhatsApp and in-app
• A published outcome split, 84 percent resolved by the agent and 16 percent escalated with context, shown on the page without a stated sample, period or method

Watch out for: the published actions are described generically, as calendar, CRM, SMS and booking. Which CRM objects and fields are connected today, and what the agent does when a write does not complete, are questions for the pilot, exactly as they are for every other row here. There is also a second surface with the same name: the reporting connector is documented as read-only by design and never changes configuration, which is the opposite of the action layer described above.

Pricing: not published. The route in is a demo and a contract, with no self-service tier.

Doesn't fit if: the phone is the only channel you use, because the value of this shape is that the call shares a profile with the messages. It also does not fit an organization with no owned channels and no opted-in contact base, one that wants public pricing and card signup, one that needs a sales CRM or a storefront and not an engagement layer, or one whose volume is low enough that a journey platform is more machinery than the problem needs.

Public evidence: the AI voice and AI agents product pages (the mid-call action and the outcome split) and the MCP integration page (the read-only reporting surface), all linked above. Checked September 2026.

What happens when the write fails?

This is the section most category comparisons leave fragmented, and it is not a small question. A write can fail while the conversation sounds perfect, which means the failure is invisible at exactly the moment someone could still fix it.

Eight questions, in the order they matter. None of them is exotic and all of them are answerable in a pilot.

• Does the platform retry, or is the retry a branch someone has to design?
• Is the operation idempotent, so that a retry updates the record it already touched?
• Does the agent get the result back in the same turn, or only a message that plays regardless?
• What does the agent say when the write does not complete, and who wrote that sentence?
• Is the record left in a pending state that something else will pick up?
• Does the call transfer to a person with the context, or does it end normally?
• Is the caller's intent preserved even when the write is lost?
• If the first object updated and the second failed, what happens to the first?

Run those against the pages and the group separates cleanly. Cognigy publishes a timeout with numbers, a retry parameter, an error threshold, branching on the error and a call failover that can transfer on a runtime error. Retell AI publishes a configurable timeout, up to five retries with exponential backoff and jitter, and the guarantee that the error text reaches the agent so the prompt can decide what to say. Vapi publishes a timeout, an optional backoff plan and the strict error contract described earlier. Thoughtly publishes the branches, and is candid that capping them is the builder's job.

The asymmetry is worth naming, because it cuts the wrong way if you leave it implicit. A platform that publishes its failure behavior is better documented than one that does not, which is not the same as being safer. What you can say is narrower and more useful: for a few platforms the public documentation answers several of the eight questions before the first meeting, and for all thirteen the pilot still has to test the complete path.

And one answer is the same for all thirteen. None of them documents platform-managed idempotency for a live CRM write on the pages checked. What exists is the requirement pushed back to you: Retell tells implementers that the receiving endpoint should be idempotent and recommends deduplicating webhooks on the event and call identifier, and Thoughtly tells webhook implementers to handle duplicate calls gracefully. Those are implementation requirements, not guarantees that the platform makes every CRM write idempotent. If your CRM has a unique-record rule that matters, enforcing it is yours, on every platform here.

🖼️ Visual suggestion, diagram: the eight failure questions as a decision strip along a single call, from the caller saying yes to the call ending, showing where each question lands. INTERNO: inserta aqui diagram.
INTERNO, brief de diseno: imagen nueva. Una banda horizontal que representa la llamada, con ocho marcas a lo largo, cada una con su pregunta en corto. Sin iconos. Azul de casa para la banda, texto en negro. Lo que tiene que quedar claro es que casi todas las preguntas caen ANTES de colgar, que es cuando todavia se pueden contestar.
Alt: The eight questions to ask about a failed CRM write, placed along the timeline of a single call from the caller's confirmation to hang-up.
Pie: Seven of the eight are answerable while the caller is still on the line, which is why they belong in the design of the call and not in the monthly report.

What should you choose if…

If the system of record is Salesforce and the priority is fewer moving parts. Salesforce Agentforce writes natively inside the platform, and the published rate card makes the cost of a voice minute and a voice action easy to model before a meeting. The caveat is in its own row: a write into any other CRM is a build.

If you have engineers and a clear workflow already drawn. Retell AI, Vapi and Bland AI all put the function in your hands, and all three publish enough pricing to model the cost in advance. Retell publishes the most about failure behavior, Vapi publishes the strictest error contract, and Bland publishes the simplest per-minute economics.

If you want the agent configured for you. Thoughtly documents mid-call CRM writes with branching and fallback. Synthflow AI ships one of the largest integration catalogs, with the caveat that its native CRM writes land at hang-up and not during the call, so anything time-critical goes through a custom action.

If the phone system is the center of the operation. Aircall, Dialpad and CloudTalk put the agent where your calls already are. Check which specific writes the agent can perform, because in all three the platform's integration count and the agent's action list are different numbers.

If governance is what has to be signed off. Cognigy publishes one of the more explicit public failure paths here, and Talkdesk publishes its structural limits up front. Both are the fastest route through a review that asks what happens on error.

If the call is one of several channels you already use. Then what decides is which one shares a customer profile with the messages you are already sending, so the call and the follow-up do not describe two different customers. That is the shape voice inside a journey is built for. If the surrounding stack is still being chosen, that is a different decision and it should be made first.

If the write does not have to happen during the call at all. Say so out loud, because it changes the shortlist and usually the price. Plenty of workflows are perfectly served by post-call logging, and paying for synchronous writes you do not need is a common way to overspend on this category.

What does the evidence actually say today?

Four kinds of claim get mixed in this category, and separating them is most of the work.

What the answer engines say is a summary of what is published, not a test of what works. Across fifty-nine live answers to this question, the sources being retrieved are mostly comparison listicles and vendor integration pages. Evidence about failures, retries and partial writes is far scarcer than evidence about capability, which is why this guide treats the failure column as a statement about pages.

What vendor pages state is the best available evidence for most rows here, and it is genuinely uneven. Some publish object names, timeout values and retry policies. Some publish the phrase "CRM integration" and a logo wall.

What a demonstration proves is narrower than either. indigitall publishes a voice flow that checks availability, reschedules the appointment and shows the record updated. The same page displays 84 percent resolved by the agent and 16 percent escalated with context, without stating the sample, the period or the method. That is a product demonstration of the mechanism, and it should be read as one.

Adjacent integration evidence, not voice write performance

The AI agents page states support cost reduction of up to 40 percent, resolution accuracy moving from 72 percent in week one to 94 percent in week twelve, and one customer reporting a 20 percent reduction in customer service operating costs. Those describe agents acting across connected systems. None of them measures a CRM write executed during a voice call, and the page does not state the sample or the period.

That describes what a connected profile does, not what a voice write does, and the loop between data, decision and action is where that architecture is described.

What only a pilot can settle is everything in the last column of the table. Object coverage, timeout behavior under your own API limits, duplicate prevention under your own matching rules, and what the agent actually says when your CRM is slow.

"What I value most is the ability to centralize the entire marketing strategy in a single environment. It allows for orchestrating campaigns across multiple channels (email, push notifications, SMS, etc.) in an automated way, based on user behavior. The segmentation is very powerful and makes it easy to work with highly specific audiences, which directly impacts campaign performance. Furthermore, real-time personalization and the activation of automated journeys make the work much more efficient." (Verified reviewer on Gartner Peer Insights)

A verified reviewer on Gartner Peer Insights wrote that, and it sits here because the reviewer is describing the profile that makes a call and a message belong to the same customer, which is the argument for putting voice inside a journey and not beside one.

"What I value most is the ease of integration, both for apps and websites. This allowed us to integrate the platform very easily." (Verified reviewer on Gartner Peer Insights)

Another verified reviewer on Gartner Peer Insights wrote the second one, and it sits here with that reviewer's own headline intact, which adds that the customer's team still has to do the integration work. That is true of every platform in this table, and it is the honest version of the phrase "easy integration".

Who is accountable for a record an agent changed?

Permissions, audit and ownership are separate from the platform choice, and they are the part that arrives late and stops a rollout.

🖼️ Visual suggestion, diagram: the chain from the protocol to the credential to the audit trail, showing which link grants the permission and which one only carries it. INTERNO: inserta aqui diagram.
INTERNO, brief de diseno: imagen nueva. Tres eslabones en linea sobre blanco, rotulados protocolo, credencial y rastro de auditoria. Solo el del centro marcado como el que concede el permiso. Sin iconos y sin candados.
Alt: The chain behind a record an agent changed: the protocol that connects, the credential that authorizes and the trail that records it.
Pie: The protocol connects, it does not authorize. The permission and the accountability both live in the credential.

The protocol does not grant the permission. MCP is named by several platforms in this category, including indigitall, and what it is is worth separating from what it promises. The specification of 18 June 2025 describes tools as the primitive that lets models "interact with external systems, such as querying databases, calling APIs, or performing computations", and it tells servers they must validate inputs and enforce access control. Everything the buyer is asking about sits one level up, in recommendations rather than requirements: the specification says there "SHOULD always be a human in the loop with the ability to deny tool invocations", and that clients "SHOULD" prompt for confirmation on sensitive operations, implement timeouts for tool calls and log tool usage for audit purposes. Read that wording carefully. MCP names how the connection is made and how a tool reports an execution error. It does not tell you whether a given surface can only read or can execute a live write, and it recommends the timeout, the confirmation and the audit log without requiring them. Those stay product decisions, and they are exactly what the table asks about.

The write runs as somebody. Every platform here executes against credentials, and those credentials carry a permission set. Bland AI states the limit directly, scoping what its agent can reach to the permissions of the connected user. Retell AI publishes what a missing permission looks like in practice, a rejection naming the scope that is absent. Deciding which objects and fields the agent may touch is a CRM administration decision, not a vendor decision, and it should be made before the first pilot call, not after the first unwanted update.

The record needs to say who changed it. A field updated by an agent and a field updated by a person look identical in most CRMs unless someone configures them not to. Timestamps, a source field and a link back to the call are the minimum, and they are what makes the difference between a support question that takes two minutes and one that takes a week.

Outbound calls carry their own record-keeping. An agent that dials a list is making a telemarketing decision before it makes a CRM decision. The Federal Trade Commission's telemarketing rules set out record-keeping and consent obligations that sit on the business placing the call, not on the vendor supplying the voice. The consent record behind the list is a thing to check before the pilot, not after it.

And the transcript is data. It contains everything the caller said, including what they said before the agent understood which record they meant. Retention, redaction and who can read it belong in the same conversation as the write, and several platforms in this table price redaction as an add-on, which is a reason to raise it early.

These questions land on different desks inside the same organization, which is why they get missed together.

FAQs: AI voice agents that update the CRM during the call

What is the difference between a CRM lookup and a CRM write-back during a call?

A lookup reads: the agent queries the CRM and uses the answer to decide what to say next. A write-back changes the record: a field is updated, a case is created, a stage is moved. Most platforms in this category do the first well, and the second is where they differ. A platform can look up a caller perfectly and still leave every change for a person to make afterwards, which is a reasonable design as long as nobody calls it automation.

Can an AI voice agent update a custom CRM field live?

Sometimes, and it depends on the route. Platforms that expose a generic function call can write anything the CRM's API accepts, including custom objects, because you define the payload. Platforms that ship native connectors are limited to the actions the connector publishes, which usually cover standard objects and may not cover yours. The practical test is to name your field in the demo and ask which of the two routes is being used.

Does a CRM integration automatically mean the agent can write records?

No, and this is the most expensive assumption in the category. The same phrase covers a post-call export, a screen pop for a human agent, a read-only reporting connector and a live write executed mid-conversation. Ask which of the four is running, and ask it about the specific field you care about, because one product can use different mechanisms for different fields.

What happens if the CRM API times out during the call?

It depends entirely on the platform, and for most of them the public pages do not say. Where it is documented, the behavior varies from the error text being handed to the agent so the prompt decides what to say, through a designed branch that offers a call-back, to the tool failing quietly and the agent moving on. That last one is the case to ask about explicitly, because it is the one the caller never notices.

Can the agent confirm a booking or an update before the CRM confirms it?

It can, and several will if nobody stops them. The safeguard is that the result of the write returns to the agent in the same turn and the agent is instructed to speak only after it arrives. A message played while the request is running is not a confirmation: it plays whether the write succeeds or fails. If the caller is going to be told something is done, the design has to make that sentence conditional on the result.

Which platforms let developers define their own CRM functions?

The developer-configurable group in this table is built around that idea, and every other group offers some version of it as an escape hatch: a generic HTTP node, a custom action, an external service action. The difference is what you get without writing one. The question to ask is how many of your required writes need a custom function, since every group here offers them in some form.

Is an MCP connector read-only or action-capable?

Both exist, and the name does not tell you which. The protocol defines tools that can call APIs and change things, and it leaves the confirmation, the timeout and the audit log to the implementation. The same vendor can ship a read-only reporting connector and an action layer that writes to live systems, both correctly described as MCP. Ask which surface is being connected, and ask what it is permitted to do.

What should a live pilot test before production?

Five things, and none of them is conversational quality. Write the correct field on an existing record, and check that a second one was not created. Confirm the result to the caller only after it lands. Fail deliberately, by pointing at a record the agent cannot update, and listen to what it says. Hit your own API limit and see what happens to the call. And read the record afterwards to check that it says an agent changed it.

---

What does a voice platform with CRM integration cost?

Three pricing shapes coexist in this table and they are not comparable line by line. Developer-configurable platforms publish per-minute rates, sometimes with a monthly platform fee and with transfers billed separately, which makes the cost easy to model and easy to underestimate, because the engineering that builds the write is not in the number. Business phone platforms price per license per month, so the agent arrives as an add-on to a seat count you already have. Enterprise suites and the CRM vendor's own agent quote, which means the number depends on volume and on what else you buy. The comparison that holds is cost per completed outcome rather than cost per minute: a cheap call that ends with a task in somebody's queue has moved the spend to payroll.

Which CRMs are actually documented for a live write?

Salesforce and HubSpot are the two that appear most often as named connectors across this table, and the CRM vendor's own agent is the case where there is no connector at all because the agent runs inside the record. Beyond those two the picture thins quickly. The reliable test is whether the vendor publishes the objects and fields the agent can write, and whether a custom field is among them, not whether your CRM appears in an integrations logo wall. Platforms that expose a generic function call can write anything the CRM's API accepts, because you define the payload. Platforms that ship a fixed action list can write what is on the list and nothing else.

How do you roll one of these out without breaking the CRM?

In an order that puts the write last, not first. Start read-only, with the agent looking up the record and using it to decide what to say, and check that it matches the right contact rather than creating a second one. Then enable a single write on a single field, on a sandbox or on a limited segment, and watch what happens when the CRM is slow. Then add the confirmation rule, so the agent speaks only after the result comes back. Only then widen the action list. The step that gets skipped is the credential review, and it is the one that stops the rollout later: the write runs as somebody, that somebody has a permission set, and somebody else has to be able to see afterwards which records an agent changed.

Why are Five9 and Genesys not in this comparison?

Both do appear in answers to this specific question. They are not rows in this comparison because the evidence set used here did not establish a verified, public path to a CRM write during the call, and listing them with an empty cell would say something about their products that their own pages do not say. Their omission is a scope and evidence decision, not a claim that they cannot perform the capability. If either is on your shortlist, apply the same four questions to it: does the agent read the right record, does it execute an authorized write at the right moment, does the result return before the agent speaks, and what happens when it does not.

Is call summarization the same as updating the CRM?

No, and conflating them is how a buying process ends up satisfied with the wrong thing. A summary is text produced after the conversation and attached to a record. An update is a field whose value changed. Most platforms in this table do summarization well and describe it prominently, because it is easier to build and it demonstrates nicely. A summary in the notes field does not move a deal stage, does not correct an address and does not create the case that triggers the next workflow. If the outcome you need is a downstream process firing, ask which field changes, not what the notes will say.

Can a small business use any of this?

Yes, and the limiting factor is rarely the price. The developer-configurable platforms are the cheapest to start and the most expensive to finish, because somebody has to build and own the write path, and in a small team that somebody usually has another job. The packaged agent builders and the business phone platforms are the realistic route: the integration arrives configured, the action list is fixed, and the ceiling is that the fixed list is what you get. The question to ask before signing is whether the one write you actually need is on that list today, because the gap between a published action and a custom one is the gap between buying and building.

About the Authors

TL;DR: There is no single best AI voice agent for outbound calls, because the AI outbound calling category splits by who builds the parts around the call. For high-volume outbound through an API, Bland AI and Retell AI publish per-minute rates and concurrency limits. For a custom build on your own stack, Vapi. For no-code campaigns on an enterprise contract, Synthflow. For teams that already run a business phone system, CloudTalk and JustCall put the agent where the calls already are. For multichannel follow-up tied to a CRM, Thoughtly. For managed enterprise and outbound contact center deployments, PolyAI, Regal and Convoso, and for compliance-heavy U.S. teams, Plura AI. indigitall places the call as a step in a customer journey, and its public page does not describe consent, Do Not Call screening or calling hours. What separates them on an AI outbound call is narrower than voice quality: who dials, to whom, under what permission, and what the platform does when nobody answers or someone asks it to stop. Several vendors' own terms place that responsibility on the customer, so the question for every one of them is which controls the product enforces and which stay with you.

An inbound support call starts because the customer picked up the phone. An outbound call starts because a system decided to dial, and almost everything that matters about an AI voice agent follows from that one difference.

Most lists of AI voice agents are written about the conversation: how natural the voice sounds, how fast it answers, how well it handles an interruption. Those differences are real, and they matter less than a set of questions that only exist when the company is the one placing the call. Who has agreed to be called. What the platform does at 9:05 p.m. in the recipient's time zone. What the agent does when it reaches voicemail, when nobody answers, or when the person says stop in the middle of a sentence.

This guide is for the people who will answer for those questions: operations and revenue leaders who run outbound, compliance and legal reviewers, and contact center managers deciding whether an AI agent takes over part of the dialing. It gives the question that separates the platforms, the three architectures the category falls into, a table built on published pages with a column for what those pages do not prove, a card for every platform, and the questions to put in writing before a pilot. It is not legal advice.

Key takeaways

• Outbound is a different product from inbound support, and the difference is permission. An inbound caller started the conversation. An outbound recipient did not, so consent has to exist before the first ring and has to be honored if the person withdraws it during the call.
• No platform makes a campaign compliant by itself. The terms of several vendors here place consent, Do Not Call screening and disclosure duties on the customer, even when a product page says calls respect those rules automatically. Both statements can be true, and the useful question is which part the product enforces.
• The controls are the comparison. Calling window, caller ID, voicemail handling, retries, Do Not Call screening, stopping during the call and human handoff decide how a campaign behaves on a bad day, and they are documented very unevenly.
• Architecture decides who builds the controls. On a developer platform they are yours to assemble. Inside a contact center or dialer built for outbound, some of them are published as product features. A journey platform needs them from somewhere else.
• "Not documented" is a statement about a page. A platform may handle more than its public pages describe, and the only way to find out is to ask, with the question written in advance.

Is there a single best AI voice agent for outbound calls?

No. The best one depends on whether the job is high-volume dialing, a custom build on your own stack, a no-code campaign, a phone system that already carries the calls, multichannel follow-up tied to a CRM, or a managed enterprise contact center.

That is also what the live answers do. The question was put to four answer engines, and sixty-eight answers were read on September 30, 2026, seventeen from each of Google AI Overviews, Perplexity, OpenAI and Copilot, with one Google answer empty. No stable ordering comes out of them. What they share is an opening line that separates outbound from inbound, a table with a "best for" column, strengths and limits, a recommendation by scenario, and a closing question about volume, call type and technical capacity.

Who comes first changes with the criterion. Bland AI leads when the criterion is volume or concurrency, Retell AI when it is balance, latency or production use, Vapi when it is developer control, and Synthflow when it is no-code. CloudTalk and Thoughtly lead on CRM and cadences, and PolyAI, Regal, Convoso and Plura AI on enterprise, contact center or regulated use.

By the count of answers that mention each platform, Retell AI appears in 64 of the 68, Bland AI in 63, Vapi in 60 and Synthflow in 56, then CloudTalk in 30, Thoughtly in 20, PolyAI and JustCall in 16, Regal in 10, Convoso in 9 and Plura AI in 8. Being mentioned often says which names a buyer will meet. It says nothing about which one fits.

So this guide does not rank. It pairs each platform with an architecture and a buyer, and it adds the column the answers leave out.

How is this comparison built?

Scope first, because a comparison that does not say what it looked at is an opinion with a table around it.

The question. "What are the best AI voice agents for outbound calls instead of inbound support", as a buyer would type it. United States, English, checked in September 2026.

What was read. Sixty-eight live answers on September 30, 2026, plus the public product pages, developer documentation, terms of service and pricing pages of every platform in the table, reopened as rendered page text on September 30, 2026.

Who is in the table. Platforms that appear in the answers with an outbound label and publish something checkable about how they place calls. CRMs that appear only as the destination of a call outcome are not rows. HubSpot and Salesforce show up in the answers that way, as the place where the record lands.

Who is not a row, and the rule behind it. A mention in a live answer is not enough to establish a product's architecture or outbound capability. The names that appear without meeting that test are listed together after the cards.

Four measurement rules, written down so anyone can rebuild the table from the vendors' own pages.

What counts as a documented control. The vendor's own page names the mechanism, for example a window setting, a screening list, a voicemail action or a stop handler. A badge, the word "compliant" or a sentence saying calls are handled automatically is recorded as a claim and kept apart from the mechanisms.

What counts as a customer responsibility. A clause in the terms or a legal page that places consent, screening, disclosure or lawful use on the customer. It is recorded next to the claim, because the two often sit on the same vendor.

What the price column means. Rates as published on September 30, 2026. A platform with no public rate is recorded as not publishing one.

What "Not documented on the pages read" means. The pages reviewed for that row do not describe the control, and nothing is concluded about the product.

indigitall is included because its public page documents outbound voice as a step in a customer journey. It is assessed with the same columns as every other platform, and any missing control remains a question for the vendor.

What is different about an outbound call compared with inbound support?

The call starts somewhere else. In inbound support the customer initiates, so the open questions are mostly about recording, disclosure and how well the agent resolves the request, the same things that make customers give up on a phone menu. In outbound the company initiates, so permission has to exist before the first ring, and every other control has to work without the recipient having asked for anything. One platform's own legal page puts the split in a sentence: the Telephone Consumer Protection Act primarily governs outbound calls and texts, and does not generally apply to communications the consumer initiates.

Nine things change when the agent dials, and each one is a question a platform either answers in writing or leaves to you. The first five have to hold before the phone rings:

• Consent. Who has agreed to be called, in what form, recorded where, and whether the platform checks that record before dialing or only assumes it.
• Who initiates. A system, a schedule or a journey trigger starts the call. The platform needs a way to show why this number was dialed at this moment.
• Calling hours and time zone. A campaign started at 2 p.m. in one state reaches recipients at different local times. The question is whether the platform enforces a window per recipient or only a start time for the batch.
• Caller ID. The number shown has to be real, consistent and recognizable, and some platforms sell a branded caller ID as a separate add-on.
• Do Not Call screening. Against which lists, how often, and whether the platform does it or expects the customer to upload a clean list.

The last four shape what happens during and after the call:

• Stopping during the call. A person can ask to stop in the middle of a sentence. The agent has to recognize it, end the call and pass the request to the system that decides the next call.
• Voicemail and answering machines. Hang up, leave a message or navigate a menu, and whether the choice is configurable.
• Retries. How many times, how far apart, and what ends the sequence.
• Handoff. When a person answers and wants a human, what travels with the transfer.

🖼️ Visual suggestion, diagram: one call on a horizontal timeline, from the decision to dial to the next action, with the nine controls placed at the moment each one has to hold, and the inbound version shown above it with fewer steps. INTERNO: inserta aquí diagram.
INTERNO, brief de diseño: imagen nueva. Dos líneas de tiempo apiladas sobre blanco, sin iconos. La de arriba es la llamada entrante, con cuatro hitos. La de abajo es la saliente, con nueve controles como etiquetas en el punto donde tienen que cumplirse. Azul de casa para la saliente y gris para la entrante. Lo que tiene que quedar claro al ojear es que la saliente tiene controles antes de que suene el teléfono.
Alt: Two timelines, an inbound call with four steps above an outbound call with nine controls placed from the decision to dial to the next action.
Pie: On an inbound call the customer has already done the hard part by calling. On an outbound call the first five controls have to hold before anyone picks up.

What the FCC said about AI voices. In a declaratory ruling adopted on February 2, 2024, and released on February 8, 2024 (FCC 24-17), the FCC stated that the restrictions the Telephone Consumer Protection Act places on an "artificial or prerecorded voice" cover technologies that generate human voices with AI. That brings the prior consent requirement to these calls, and prior express written consent when the call introduces an advertisement or contains telemarketing. The text is on the FCC's document server, and the FCC's own summary is shorter.

What the FTC says about timing, lists and opt-outs. The FTC's guidance on the Telemarketing Sales Rule sets calling-time restrictions, including a general prohibition on calls before 8 a.m. or after 9 p.m. in the recipient's time zone, and separately addresses the National Do Not Call Registry and the seller's own list, the automated opt-out required for prerecorded telemarketing calls, the ban on cold prerecorded calls and the transmission of caller ID. The category of call and any exemption still need to be checked for each campaign.

What is still moving. The rules on revoking consent are being adjusted. On January 6, 2026, the FCC (DA 26-12) extended until January 31, 2027, the waiver of the requirement that a revocation made in response to one type of informational message apply to all future communications. An FCC fact sheet dated September 9, 2026 describes changes that would let a revocation apply only to the category of message and would set an exclusive means of revocation. All of this is as of September 30, 2026. A rule that keeps changing is a reason to ask a vendor how the product follows it.

Which category a given call falls into, a reminder or a marketing call, is a decision for the company and its counsel. What a comparison can do is show which platforms publish how the product handles the mechanics, and which leave them to the customer.

What do the live answers compare, and what do they skip?

They compare the controls as side notes. Of the sixty-eight answers, 25 mention the TCPA, 23 mention Do Not Call, 19 mention consent, 18 mention human handoff, 10 mention voicemail or an answering machine, and 7 mention calling hours, calling windows, time-zone scheduling or related timing rules, all of them from OpenAI.

None of them has a matrix of which platform documents which control, or a comparison of what happens when a person, an answering machine or nobody answers. None treats permission, timing, recipient, outcome and next action as the unit of a call. One OpenAI answer comes closest with a single sentence: the platform may provide controls, but it does not make the campaign compliant by itself.

That gap is why this page exists, and why the table below ends with a column about what a platform's published pages do not prove.

Which architectures exist for AI outbound calling agents?

Three, and the architecture decides who builds the controls around an outbound AI voice agent.

A voice agent or infrastructure platform built for developers. Vapi, Retell AI and Bland AI sit here, with Bland AI positioned for outbound at scale, and Synthflow as the no-code builder of the same kind. You assemble the campaign: the list, the schedule, the number, the screening and the voicemail behavior. The platform gives you the parts, and the terms say whose job it is to use them lawfully.

A phone system, communications platform or contact center with an AI agent. CloudTalk, JustCall, Plura AI, Regal, Convoso and PolyAI. The agent is part of a wider communications or dialing operation. Some of these publish outbound compliance features as part of the dialer, while others publish workflows, concurrency or communications channels without documenting the controls that run before the AI dials.

A CRM or customer journey with a voice step. Thoughtly, with voice, SMS and email inside CRM flows, and indigitall, where the call is one channel in a journey. The call is triggered by what the customer did.

🖼️ Visual suggestion, diagram: the three architectures as three columns, each showing what the platform provides and what the buyer has to bring, with the platforms placed under their column: Bland AI, Retell AI, Vapi and Synthflow under the first, CloudTalk, JustCall, Plura AI, Regal, Convoso and PolyAI under the second, Thoughtly and indigitall under the third. INTERNO: inserta aquí diagram.
INTERNO, brief de diseño: imagen nueva. Tres columnas sobre blanco, sin iconos. Cada una con dos bloques, "lo que trae" y "lo que aportas tú", y los nombres de plataforma debajo. Azul de casa para los títulos y gris para los bloques. Lo que tiene que quedar claro al ojear es que la diferencia entre columnas es quién construye los controles, no quién habla mejor.
Alt: Three columns showing the architectures for an outbound AI voice agent, each with what the platform provides, what the buyer supplies and its platforms.
Pie: The columns differ in who builds the controls. A platform can sound excellent in any of them and still leave the calling window to you.

The comparison table

When shortlisting an AI agent for outbound calls, read the last two columns first. The controls column is where vendors differ most, and the last column is what decides the pilot.

One row has no engine label, and that is deliberate. indigitall does not appear in the live answers to this question, so there is no wording from an engine to put in that cell. Writing one would falsify the whole column, so the cell says what it is.

Each card says where its cells come from. It names the public pages used to build the comparison and the date they were checked, and its claims stay within what those pages document. When a control is not described, the card says so and leaves it as a question for the vendor.

The last column describes published pages. "Not documented on the pages read" means the pages reviewed for that row did not describe the control. It does not show that the product cannot perform it, and a platform may handle consent, hours and screening in ways it has not published. The only way to find out is to ask, with the question already written, which is what the closing sections are for.

The platforms, one by one

The twelve platforms are grouped by architecture, in the order of the table, and each card draws only on the platform's own public pages for what that platform does. The wider category they sit in is mapped in the guide to AI agent software. The groups describe who builds the controls, which is the decision underneath the others, and they rank nothing.

Bland AI

A programmable voice platform with an API, batch endpoints and self-serve plans, positioned for outbound at volume.

🖼️ Visual suggestion, screenshot [external]: the Bland AI page at https://www.bland.ai/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://www.bland.ai/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: Bland AI home page with a headline on voice AI for regulated industries, and the latency figure and compliance badges below.
Pie: It sells infrastructure and control, and the calling rules are yours to apply. Bland AI.

Best for: high-volume outbound campaigns, the label the answers repeat most for it.

What it does well: it publishes its economics and its limits on one page. Per-minute rates, daily and hourly call caps and concurrency for each plan sit side by side, so volume can be modeled before a conversation with sales. Batches accept many call objects, each with its own start time.

Key features:

• A batch endpoint that takes a list of call objects, each with a start time
• Daily caps of 100 calls on Start and 2,000 on Build, and hourly caps of 100 and 1,000
• 10 concurrent calls on Start, 50 on Build and custom on Enterprise
• A separate transfer rate per minute for calls handed to a person

Watch out for: the terms make the customer solely responsible for lawful use, in general terms, and the pages read describe no consent, Do Not Call or calling-window mechanism. The batch example also treats the time zone of a start time as optional, defaulting to UTC, so a scheduled start that leaves it out is read in UTC.

Pricing: published. 0.14 US dollars per minute on Start with no platform fee, and 0.12 plus a 299 US dollar monthly platform fee on Build. Transfers are priced at 0.05 and 0.04 per minute. Enterprise is custom. Rates as published on September 30, 2026.

Doesn't fit if: you need more than 2,000 calls a day without an enterprise contract, because the top self-serve plan caps there.

What the source does not prove: that a scheduled start respects the recipient's local time, or how voicemail and a request to stop are handled, which the pages read do not describe.

Public evidence: the Bland AI pricing page, the batch create API reference and the terms of service. Checked September 30, 2026.

Retell AI

Voice infrastructure sold to developers, with batch calls, voicemail handling and a published concurrency model.

🖼️ Visual suggestion, screenshot [external]: the Retell AI page at https://www.retellai.com/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://www.retellai.com/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. La página redirige al español según el idioma del navegador: captura la versión en inglés. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: Retell AI home page headlined as the number one AI voice agent platform for automating phone calls, with the drag and drop canvas shown.
Pie: The product is sold as parts a team assembles, and the page leads with the build and leaves the rules of the call to other pages. Retell AI.

Best for: best overall balance and low-latency outbound campaigns, in the engines' words.

What it does well: the batch page covers the operating parts of a campaign in one place. A CSV carries per-recipient columns, the agent has a from number, time windows are configured for the batch, and a reserved concurrency setting keeps a batch from taking every line. Voicemail detection lets the agent hang up, leave a message or navigate a menu, and it can be tested without dialing by playing a recording into a web call.

Key features:

• CSV batch calls where extra columns reach the agent as variables
• Time windows configured for each batch
• Reserved concurrency for other calls, so a batch does not take every line
• Voicemail and IVR detection with hang up, leave a message or navigate the menu
• A branded call ID as a paid add-on

Watch out for: the terms place compliance with telemarketing and recording laws on the customer and require screening against Do Not Call lists at least every 31 days. The batch page does not say whether the time windows follow each recipient's time zone.

Pricing: published. Voice agents from 0.07 to 0.31 US dollars per minute depending on the voice and model, with 20 concurrent calls included. Batch calls add 0.005 per dial and a branded call ID adds 0.10 per outbound call. Rates as published on September 30, 2026.

Doesn't fit if: you want a finished campaign manager that applies your compliance rules for you, because the terms leave those checks with the customer.

What the source does not prove: that time windows follow each recipient's time zone, and how consent records are kept.

Public evidence: the Retell AI pricing page, the batch call and voicemail documentation and the terms of service. Checked September 30, 2026.

Vapi

A platform for developers creating voice agents, with campaigns built from a caller, a number, contacts and a schedule.

🖼️ Visual suggestion, screenshot [external]: the Vapi page at https://vapi.ai/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://vapi.ai/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: Vapi home page headlined "Voice agents for builders", with the developer entry points listed.
Pie: The page addresses builders, and builders are where the campaign controls come from here. Vapi.

Best for: custom and engineering-heavy outbound systems, and developer-led outbound.

What it does well: a campaign is made of a caller, an outbound-capable number, contacts and a schedule that sets when dialing starts, when it must stop and how many calls run at once. Contacts begin pending and every attempt has a result to review in one place. The outbound documentation is also direct about the stakes: it says automated dialing without consent breaks FCC law, and it points to its own consent guide.

Key features:

• Campaigns that call a list with an assistant or a squad
• A schedule with a start, a stop and a concurrency limit
• Values per contact to personalize each conversation
• An imported outbound-capable number, since free Vapi numbers are not supported for campaigns
• A result for every attempt, reviewable in the campaign

Watch out for: no consent-capture feature appears on the pages read, and the stop time is a campaign setting, not a per-recipient one. Models, transcription and voice are billed at provider cost on top of the hosting fee, and outbound needs a number imported from a telephony provider.

Pricing: published. A pre-paid hosting fee of 0.05 US dollars per minute with models passed through at cost. The free tier allows 4 concurrent calls, a Core plan at 29 US dollars allows 10, and a Pro plan at a 999 US dollar monthly minimum allows 30. Rates as published on September 30, 2026.

Doesn't fit if: you want one invoice for the whole call, because hosting, model, transcription and voice are separate lines.

What the source does not prove: how consent gets recorded, and whether the stop time works per recipient.

Public evidence: the Vapi pricing page and the outbound calling and campaigns documentation. Checked September 30, 2026.

Synthflow

A no-code voice agent platform sold on enterprise contracts, with outbound campaigns started through an API.

🖼️ Visual suggestion, screenshot [external]: the Synthflow page at https://synthflow.ai/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://synthflow.ai/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: Synthflow home page with a headline on enterprise-ready voice agents for automated phone calls, and the integration catalog below.
Pie: The promise is a working agent without engineering, and the pricing page is where the enterprise model shows. Synthflow.

Best for: no-code outbound workflows, for nontechnical sales and operations teams.

What it does well: a campaign is a list, an agent and the hours you allow. It dials through the list as fast as the account's concurrency permits. Every dialed recipient becomes a regular call with transcripts, recordings and webhooks, and a campaign can be paused, resumed, canceled and extended with more recipients.

Key features:

• A campaign created with one request to the Platform API, returning an id to follow it
• Dialing inside the hours you allow
• Pause, resume, cancel and add recipients later
• Every call in the same history, with transcripts and recordings

Watch out for: campaigns are documented on Enterprise plans and activated through support, which makes the no-code builder a platform whose campaigns are started through an API. The transparency statement leaves the AI disclosure to the customer to configure in line with the law that applies.

Pricing: published as a floor. Enterprise contracts start at 30,000 US dollars annually, and the final price depends on volume, concurrency, telephony setup, integrations, security and launch support. As published on September 30, 2026.

Doesn't fit if: you want to start small and self-serve, because campaigns sit on the Enterprise plan.

What the source does not prove: how the allowed hours relate to each recipient's time zone. The transparency statement has a section on consent and legal compliance for outbound AI calls, and it is the first page to read before a pilot.

Public evidence: the Synthflow pricing page, the campaigns documentation and the AI transparency statement. Checked September 30, 2026.

Thoughtly

An engagement platform with voice, SMS and email agents inside CRM-connected flows.

🖼️ Visual suggestion, screenshot [external]: the Thoughtly page at https://thoughtly.com/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://thoughtly.com/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: Thoughtly home page headlined that everyone deserves a follow up, with customer logos below.
Pie: The page sells the follow-up, and the outbound call is one of the ways it happens. Thoughtly.

Best for: multichannel cadence, automated follow-up and speed-to-lead follow-up.

What it does well: the call is part of a cadence. Outbound runs from a CSV upload. Voice, SMS and email sit in the same workflows, and two-way CRM sync with HubSpot, Salesforce and Zoho, among others, returns the outcome of a call to the record.

Key features:

• Bulk upload from CSV to run outbound calls
• Voice, SMS and email in the same workflows
• Two-way CRM sync, with HubSpot, Salesforce and Zoho among the named systems
• 200 or more integrations and a REST API
• A quoted tier with 100 or more concurrent calls and a TCPA review

Watch out for: the terms make compliance the customer's sole responsibility, name the TCPA, the Telemarketing Sales Rule and Do Not Call rules, and require a truthful caller ID and the legal disclosures. The Flex plan is described for inbound conversion and lists 10 concurrent calls.

Pricing: partly published. Flex starts at 500 US dollars a month and the Tailored tier is a quote. No per-minute rate appears on the pricing page. As published on September 30, 2026.

Doesn't fit if: outbound volume is the main job and you want published per-minute rates.

What the source does not prove: what a campaign enforces before dialing, since the pages read show the upload and the terms but not the controls.

Public evidence: the Thoughtly pricing page, the terms of service and the bulk upload documentation. Checked September 30, 2026.

CloudTalk

A business phone system with AI voice agents that handle inbound and outbound calls.

🖼️ Visual suggestion, screenshot [external]: the CloudTalk page at https://www.cloudtalk.io/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://www.cloudtalk.io/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: CloudTalk home page headlined "Every call answered, 24/7. By your team, or by AI.", with the call handling blocks below.
Pie: Answering is the promise on the page, and outbound sits one click away in the agent product. CloudTalk.

Best for: sales teams wanting a phone system plus AI, and small and mid-size companies that want the phone system integrated.

What it does well: the agent runs inside a system that already has numbers, users and CRM sync. The agent page says it handles inbound and outbound calls in more than 70 languages and accents without code, and bundles of voice agent minutes are priced separately from the seats.

Key features:

• AI voice agents for inbound and outbound calls in 70 or more languages and accents
• A business phone system that syncs with your CRM
• A branded caller ID as an add-on
• Bundles of voice agent minutes starting at 200 a month

Watch out for: prices are shown in euros, with seat plans billed per user and per year, and the pages read do not describe consent, Do Not Call screening or calling hours for the AI agent.

Pricing: published. Seat plans start at 25 euros per user per month on Starter and 29 on Essential, billed annually, with a minimum of one license. Voice agent bundles start at 99 euros for 200 minutes a month. As published on September 30, 2026.

Doesn't fit if: you have no use for a business phone system, because the AI agent is positioned as part of a business phone system rather than as a standalone voice agent.

What the source does not prove: which outbound controls the AI agent uses and which belong to the human dialer, and what happens when the agent reaches voicemail.

Public evidence: the CloudTalk pricing page and the AI voice agents page. Checked September 30, 2026.

JustCall

A business phone system and dialer with an outbound AI voice agent as a product of its own.

🖼️ Visual suggestion, screenshot [external]: the JustCall outbound AI voice agent page at https://justcall.io/product/outbound-ai-voice-agent/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://justcall.io/product/outbound-ai-voice-agent/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: JustCall outbound AI voice agent page, with the headline about outbound calls and the problem and solution pairs listed below.
Pie: The page states the compliance promise as one line next to the problem it answers, which is where a buyer should start asking. JustCall.

Best for: outbound calling inside a full phone system, and lead qualification plus booking.

What it does well: it states the compliance promise in the plainest words of any vendor here. The product page says every call respects consent, time-of-day and Do Not Call rules automatically. It also says call summaries, outcomes and next steps sync to the CRM in real time.

Key features:

• An outbound AI voice agent with its own product page
• Automatic respect for consent, time-of-day and Do Not Call rules, as the page states it
• Call summary, outcome and next steps synced to the CRM
• Business phone system, messaging and call center software on the same platform

Watch out for: the page also says the agent is only available for compliant use cases, that every call trigger includes built-in consent and that the agent calls only opted-in contacts. It still does not describe the mechanism, where the list comes from or how a stop during the call is handled. Ask what "automatically" covers and which consent record the product checks.

Pricing: partly published. Business phone plans start at 29 US dollars a month according to the pricing page, and the AI voice agent is listed there as an add-on. As published on September 30, 2026.

Doesn't fit if: you want only an agent and not a phone system, because the engines describe it as outbound calling inside a full phone system.

What the source does not prove: what "automatically" covers, who supplies the consent record, and whether the same controls apply to AI calls and to human ones.

Public evidence: the JustCall outbound AI voice agent page and the pricing page. Checked September 30, 2026.

PolyAI

An enterprise conversational AI vendor with a documented API for placing outbound calls.

🖼️ Visual suggestion, screenshot [external]: the PolyAI page at https://poly.ai/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://poly.ai/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: PolyAI home page headlined "Own every conversation", with the enterprise dialog agent positioning below it.
Pie: The page sells the conversation and says little about the outbound call, which is why the API reference is the better source. PolyAI.

Best for: enterprise managed deployments and complex outbound conversations.

What it does well: the public API reference includes a call that places a programmatic outbound call to a target number and tracks its status, next to conversation, handoff, messaging, SMS and webhook APIs. That is concrete, checkable outbound documentation from an enterprise vendor.

Key features:

• An API that triggers an outbound call and tracks its status
• Handoff, messaging, SMS and webhook APIs in the same reference
• Keys provisioned to a region, with a different host for each API family

Watch out for: the pages read describe a single-call API and not a campaign manager, and the product is sold through a demo request.

Pricing: no rate appears on the pages read. The documentation offers a demo request.

Doesn't fit if: you want to self-serve a campaign, because the positioning the engines repeat is managed enterprise deployments.

What the source does not prove: campaign controls, calling windows, consent or Do Not Call handling.

Public evidence: the PolyAI API reference. Checked September 30, 2026.

Regal

An AI agent platform for contact centers, with agents that staff progressive and predictive dialer campaigns.

🖼️ Visual suggestion, screenshot [external]: the Regal page at https://www.regal.ai/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://www.regal.ai/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: Regal home page addressed to customer experience leaders, with the agent platform layers and the customer logos below.
Pie: The page names a safety layer for compliance without describing what it enforces, which is the question the last column of the table carries. Regal.

Best for: B2C outbound sales and collections, and regulated industries, as the engines label it.

What it does well: it is built around the dialer. A post from the vendor, last updated on March 17, 2026, says its AI agents now staff progressive and predictive dialer campaigns, reducing time lost to rings and voicemails. The home page names a layer of built-in safeguards for privacy, compliance, hallucination control and escalation.

Key features:

• AI agents that staff progressive and predictive dialer campaigns
• A safety layer for privacy, compliance, hallucination control and escalation, as the home page names it
• A business context layer that draws on customer data, knowledge and workflows
• A vendor guide titled Outbound Campaign Compliance 101, which names the Telemarketing Sales Rule, the TCPA and Do Not Call

Watch out for: the compliance material is a named layer and a set of vendor posts, with no mechanism described on the pages read. Ask which checks run before the AI agent dials.

Pricing: no rate appears on the pages read.

Doesn't fit if: your use case is inbound support, because the engines place it in outbound sales and collections.

What the source does not prove: which checks run before an AI agent dials, and whether the safety layer enforces a calling window.

Public evidence: the Regal home page and its blog post on progressive dialing for AI agents. Checked September 30, 2026.

Convoso

An outbound contact center and dialing platform that offers a voice AI agent to qualify leads.

🖼️ Visual suggestion, screenshot [external]: the Convoso voice AI agent page at https://www.convoso.com/voice-ai-agent/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://www.convoso.com/voice-ai-agent/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: Convoso voice AI agent page with a headline on qualifying more leads and giving sellers more conversations that can close.
Pie: The agent is presented as part of the dialing platform, and the compliance controls are described on another page. Convoso.

Best for: compliance-heavy industries and contact-center scale, per the engines.

What it does well: it lists compliance among its product features. A compliance page describes state dialing rules tracked against each lead record, Do Not Call screening against internal, federal and state lists, and a reassigned numbers database. The voice agent qualifies leads at scale and hands sales-ready prospects to sellers.

Key features:

• State dialing rules tracked per lead record
• Do Not Call screening against internal, federal and state lists
• A reassigned numbers database
• A voice AI agent that qualifies leads and hands them to sellers

Watch out for: the compliance page describes the platform and does not say whether the voice agent runs inside those controls. A response-time figure on the agent page is a vendor claim and is not repeated here.

Pricing: by request. The site offers a path to pricing and no public rate.

Doesn't fit if: you want a standalone voice agent without a dialing platform, because the agent is presented as part of Convoso's outbound platform.

What the source does not prove: whether the compliance controls apply to calls the voice agent places, and how the agent handles voicemail.

Public evidence: the Convoso compliance page and the voice AI agent page. Checked September 30, 2026.

Plura AI

A communications platform with AI agents for voice, SMS and RCS, sold on annual contracts.

🖼️ Visual suggestion, screenshot [external]: the Plura AI page at https://www.plura.ai/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://www.plura.ai/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: Plura AI home page with a headline on AI agents for voice, SMS and RCS, and the communications platform positioning below.
Pie: The page presents a communications platform, and the plan limits on the pricing page say more about outbound than the home page does. Plura AI.

Best for: enterprise compliance-heavy teams, in the engines' words.

What it does well: its plan limits are public, so a buyer can test them. The Multi plan lists 5 concurrent outbound calls and 8 concurrent inbound calls, unlimited workflows, and caller ID monitoring among its features.

Key features:

• Voice, SMS and RCS agents on one platform
• Separate concurrency limits for outbound (5) and inbound (8) calls on the Multi plan
• Unlimited workflows
• Caller ID monitoring listed among the plan features

Watch out for: a glossary page on its site says the company operates as an FCC-licensed carrier. It is the vendor's claim about itself. The pages read give no license number, and the claim does not appear on a product page, so it stays a question for the vendor. Plans are annual contracts.

Pricing: published. The Multi plan is 5,000 US dollars a month plus overages on a yearly contract billed monthly, with agent talk time at 15 US dollars an hour, transfer time at 0.02 per minute, SMS at 0.06 per message and phone numbers at 3 US dollars a month. As published on September 30, 2026.

Doesn't fit if: you need month-to-month flexibility or a small pilot, because the published plans start at 5,000 US dollars a month on annual contracts.

What the source does not prove: the license claim, and how campaigns, calling windows and consent are handled, which the pages read do not describe.

Public evidence: the Plura AI pricing page and a glossary page on licensing. Checked September 30, 2026.

indigitall

Outbound voice as one step inside a customer journey.

🖼️ Visual suggestion, screenshot [external]: the indigitall page at https://indigitall.com/en/ai-voice/, framed on the top third. INTERNO: inserta aquí screenshot.
INTERNO, brief de diseño: captura externa de https://indigitall.com/en/ai-voice/. Encuadre: tercio superior, ancho completo de navegador de escritorio, sin barra de marcadores y sin banner de cookies. Al montar, añade al final del pie el día de la captura y la fuente, "capturada el" más la fecha.
Alt: indigitall AI Voice page showing the outbound call as a step inside a customer journey, with the trigger and the actions listed.
Pie: The call is drawn with a trigger before it and an action after it, which is the shape this section describes. indigitall.

Best for: organizations already running reminders and follow-ups over several channels that want the call to share the same customer profile and the same rules. There is no engine label for indigitall, so this is an editorial fit.

What it does well: the trigger comes before the call. A journey step or a churn or no-show prediction triggers it, and the page describes it as a next best action placed automatically when a text has not landed, not a blast to a list. It documents three uses, retention when email and WhatsApp do not convert, operations such as confirming or rescheduling an appointment and reducing no-shows, and proactive alerts such as renewals or a delivery change.

Key features:

• Outbound calls triggered by a journey step or a churn or no-show prediction
• Actions on connected systems through MCP, listed as calendar, CRM, SMS and booking
• A handoff to a person that carries the transcript, the detected intent and the steps already taken
• The same customer profile across push, email, SMS, WhatsApp and in-app
• A published outcome split of 84 percent resolved by the agent and 16 percent escalated with context, shown without a sample, period or method and not stated as a measure of outbound calls

Watch out for: the page does not describe consent, Do Not Call screening, calling hours, AI disclosure or voicemail detection, and those questions apply to indigitall exactly as they do to every other row. It also publishes no concurrency, latency, answer rate or price. There is a second surface with the same family name: the reporting connector is documented as read-only by design, which is the opposite of the actions described above.

Pricing: not published. The route in is a demo and a contract, with no self-service tier.

Doesn't fit if: the phone is the only channel you use, because the value of this shape is that the call shares a profile with the messages. It also does not fit an organization with no owned channels and no opted-in contact base, one that wants a high-volume dialer with campaign tooling built for outbound, one that wants public pricing and card signup, or one that needs a sales CRM or a storefront and not an engagement platform.

What the source does not prove: that consent, calling hours and screening are handled, that the 84 percent applies to outbound, or that any customer runs outbound voice today, since no published case covers it.

Public evidence: the AI voice and AI agents product pages and the MCP integration page, all linked above. Checked September 30, 2026.

Which other names show up in the answers?

Several names appear in the answers, but no public page that was read establishes their architecture or outbound capability. Under the rule that a mention is not enough, they get no card. Goodcall, 11x, Aloware, Cognigy, Replicant and Observe.AI appear as agents, contact center software or analytics, in answers that do not agree on what they are for. Plivo, Telnyx, SignalWire and LiveKit appear as communications infrastructure, the part under an agent rather than the agent itself. A long tail of single mentions, among them AiSA-X, Leaping AI, Brilo.ai, Air AI, LuMay, Orum, Sierra, Regie.ai and Commplify, appears once or twice each. A name in this paragraph shows how much of the answers' evidence could be checked on a public page, and tells you nothing about the product.

Can AI agents make outbound calls within the rules? What to ask a vendor about consent, Do Not Call and calling windows

Put these in writing before a pilot, and ask every vendor on the shortlist the same ten, including the ones whose pages sound reassuring. The answers decide more than any demo. The first five apply before the first ring.

1. Who records consent, and does the platform check it before dialing? Ask for the field where the consent record lives and for what blocks a call when it is missing. A platform that assumes the list is clean and a platform that refuses to dial a number without a record are different products, and the pages read rarely say which one it is.

2. Which Do Not Call lists does it screen, and how often? One vendor's terms require screening at least every 31 days and leave the job to the customer. Another publishes screening against internal, federal and state lists as a feature. Ask whether the product does it or expects you to upload a clean list.

3. Whose time zone sets the calling window? The FTC guidance sets a general prohibition on calls before 8 a.m. or after 9 p.m. in the recipient's time zone, subject to the category of call and any exemption. Ask whether the window is enforced for each recipient or set once for the batch, and what happens to the calls still queued when the window closes.

4. What does the recipient see as the caller ID? A consistent, recognizable number is a requirement in the FTC guidance, and some platforms sell a branded caller ID as a separate add-on. Ask who is responsible for the number's reputation.

5. What does the agent say first? One vendor's transparency statement leaves the AI disclosure to the customer to configure. Ask where the disclosure is set, what the default is and who can change it.

The last five cover the call itself, the contract and the test.

6. What happens when the person says stop? A recipient can withdraw consent by any reasonable means, and the campaign must recognize and process it during the call. Ask how the agent detects it, how fast it reaches the list and how the vendor follows the revocation rules that are changing in 2026.

7. What happens at voicemail and when nobody answers? One documentation page lets the agent hang up, leave a message or navigate a menu. Ask for the cap on retries, the spacing between them and what ends the sequence.

8. What travels with the handoff? When a person answers and asks for a human, ask whether the transfer carries the transcript, the detected intent and the steps already taken, and what the agent does if nobody is free.

9. Which part is the product and which part is the customer? Set the product page and the terms side by side. A page that says calls respect the rules automatically and a clause that makes compliance the customer's sole responsibility are both on record at the same vendor, and you want to know what each covers.

10. What can be tested without dialing a real person? One documentation page describes testing voicemail behavior by playing a recording into a web call. Ask for a test that exercises each control above with a list of numbers that have consented.

Who is responsible when an AI outbound call breaks a rule?

Two statements sit side by side in the vendors' pages. A product page says every call respects consent, time-of-day and Do Not Call rules automatically. The terms of other vendors say compliance is the customer's sole responsibility, name the laws, and put the duty to screen, to obtain consent and to disclose on the customer. Several of these clauses are general, as in a platform that makes the customer solely responsible for lawful use.

Neither rules out the other. A platform can enforce a calling window and still leave the consent record to you. The split decides the risk: which controls the product applies on its own, which it offers as settings, and which it leaves entirely to the customer. One engine answer puts it in a line: the platform may provide controls but does not make the campaign compliant by itself.

That is also why a platform that publishes little is not automatically worse, and one that publishes a lot is not automatically safer. The pages show what a vendor chose to write down. The contract and a pilot show what the product does.

What should you choose if…

If the job is high-volume outbound and you have engineers. Bland AI publishes caps, rates and concurrency for each plan, and Retell AI publishes the most about batch windows, reserved concurrency and voicemail handling. Both leave the calling rules with you.

If you want full control of the stack. Vapi builds a campaign from parts: caller, number, contacts and schedule. The cost is an imported number and separate bills for the model, the transcription and the voice.

If you have no developers and an enterprise contract is acceptable. Synthflow is the no-code answer, with campaigns started through an API and activated through support. If the contract is the obstacle, the phone-system route below is the alternative.

If a business phone system already carries your calls. CloudTalk and JustCall put the agent where the calls already are. Check which outbound controls the agent uses and which belong to the human dialer, and ask JustCall what "automatically" covers.

If the work is follow-up across channels tied to a CRM, and consent is already captured. Thoughtly fits cadences that mix voice, SMS and email, the kind of sequence described in how a nudge recovers abandoned applications, with the terms making compliance yours and a TCPA review on the quoted tier.

If dialing compliance built into the platform is the requirement. Convoso publishes the most about screening and state rules, and Regal and PolyAI serve managed enterprise deployments. In all three, ask whether the voice agent runs inside the same controls as the human dialer.

If compliance-heavy U.S. operations need a communications platform on an annual contract. Plura AI publishes its plan limits, and its licensing claim is a question to put to the vendor.

If the call is one channel among several you already run. Then choose the platform that shares a customer profile with the messages you already send, so the call and the follow-up describe the same customer. That is the shape voice inside a journey is built for, and the same questions above apply to it. If the surrounding stack is still being chosen, that is a different decision.

If a person should make the first call and the agent should handle only what follows. Say so out loud, because it changes the shortlist and usually the permission question. Plenty of workflows are served by a confirmation or a reminder sequence that never needs an automated outbound voice call.

FAQs: AI voice agents for outbound calls

Can an AI voice agent legally make outbound calls in the United States?

The FCC ruled in February 2024 that the Telephone Consumer Protection Act's rules on an artificial or prerecorded voice cover AI technologies that generate human voices. That means prior consent, and prior express written consent when the call contains advertising or telemarketing. Whether a particular call is that kind of call is a question for your counsel, and this guide is not legal advice.

What is the difference between an AI voice agent for outbound calls and one for inbound support?

In inbound support the customer starts the call, so the questions are about resolution, recording and disclosure. In outbound the company starts it, so permission, calling hours, Do Not Call screening, caller ID, voicemail, retries and stopping during the call all have to work before anyone has asked for anything.

What is the best AI voice agent for outbound calls?

There is no single best one. The answers that were read lead with Bland AI for volume, Retell AI for balance, Vapi for developer control, Synthflow for no-code, CloudTalk and Thoughtly for CRM and cadences, and PolyAI, Regal, Convoso and Plura AI for enterprise and regulated use. What fits depends on volume, call type and how much you want to build.

Does a reminder call need consent?

The FCC distinguishes calls that contain advertising or telemarketing, which need prior express written consent, from calls that do not. An AI-generated voice is treated as an artificial voice in both cases. Which type a reminder is depends on its content and context, and the company decides that with its legal counsel.

What happens when the agent reaches voicemail?

It depends on the platform. One documentation page lets the agent hang up, leave a message or navigate a phone menu, and lets you test the behavior without dialing. For the others, the pages read do not describe it, so it is a question to ask.

Can the person stop the calls during the conversation?

A recipient can revoke consent by any reasonable means, including saying so during a call, according to one platform's legal page. The campaign must recognize and process the request. How quickly the stop reaches the list, and how revocation rules apply after the 2026 changes, are questions for the vendor.

Do I need my own phone number?

On some platforms, yes. Vapi does not support its free numbers for outbound campaigns and requires an imported number, and Retell AI asks you to select a from number bound to the agent. A business phone system such as CloudTalk or JustCall brings its own numbers.

What does an AI voice agent for outbound calls cost?

On September 30, 2026, published rates range widely. Bland AI lists 0.14 and 0.12 US dollars per minute on its two self-serve plans, Retell AI 0.07 to 0.31 per minute, and Vapi 0.05 per minute of hosting plus model and voice costs at provider cost. CloudTalk bundles start at 99 euros for 200 minutes, Plura AI's Multi plan is 5,000 US dollars a month, and Synthflow's enterprise contracts start at 30,000 US dollars a year. PolyAI, Regal and Convoso do not publish a rate on the pages read, and neither does indigitall.

Which platform fits a small team with no developers?

The answers point to Synthflow for no-code campaigns, though its campaigns sit on an Enterprise plan. A business phone system such as CloudTalk or JustCall is the other route, with the agent inside a product that already has numbers and CRM sync.

Can indigitall place outbound voice calls?

Its AI voice page describes outbound conversational voice as a channel in the customer journey, triggered by a journey step or a churn or no-show prediction. The page does not describe consent, Do Not Call screening, calling hours, AI disclosure or voicemail detection, and it publishes no price. Those are the first things to ask, as with any vendor here.

Why are some platforms not in the comparison?

A mention in a live answer is not enough to establish a product's architecture or outbound capability. Names that appear without a public page that shows how they place calls are listed without a card.

How should you test a platform before a full campaign?

Start with a small list of numbers that have consented, run every control from the ten questions above against it, and check the results in the platform's own call history. Where the documentation allows it, test voicemail and menu behavior without dialing, and read the terms next to the product page before the first real call.

About the Authors