The journey places the call.
A step in the Customer Journey triggers an outbound call. The right call, to the right customer, at the right moment. Not a blast to a list. A next best action, placed automatically when text hasn't landed.
I am indigitall AI, how can I help you today?
Resources
eBooksFree ebooks, whitepapers...
BlogStories, insights and advice
indigitall MCP NewConnect Your Platform Stats to Any AI Agent
IntegrationsHubspot, Salesforce & more
DocumentationComplete guide to indigitall
Outbound conversational voice, added to the Customer Journey like any other channel. The AI calls, understands, acts on your systems in real time, and hands off to a human when it matters, triggered by customer behavior or a journey step. Not a menu bot. A call that actually converses.
A step in the Customer Journey triggers an outbound call. The right call, to the right customer, at the right moment. Not a blast to a list. A next best action, placed automatically when text hasn't landed.
Natural, real-time voice understands intent: it is not a tone menu. It answers questions and acts on your live systems: it books the appointment, confirms the order, resolves the query and updates the record while still on the call.
When a call needs a person, AI Voice hands it off to a human agent with the full transcript and the customer's context, so the customer never repeats themselves. No cold transfers.
Runs 24/7 and scales to thousands of calls at once. Every call feeds the journey, so the next touch is smarter. No queue, no hold music, no human required to keep it running.
Powering millions of omnichannel interactions every single month.
The indigitall AI Voice Agent is not an IVR with a nicer voice. It's an AI agent connected to your systems, holding a natural phone conversation in your customer's language and taking action while they're still on the line. Here's what it sounds like in production.
Real scenarios where a live voice call closes what text can't: triggered automatically, resolved end-to-end, escalated to a human only when it matters.
When email and WhatsApp don't convert, the journey escalates to an AI call that picks up right where the customer left off and closes by voice.
An AI call confirms or reschedules the appointment, cuts no-shows and updates your system, with no agent on the line and no hold music.
Trigger a friendly voice call for a renewal, a delivery change or a time-sensitive alert, at the step of the journey where it matters most.
G2 is the world's largest B2B software review platform, with 90 million verified reviews. These 76 badges were earned from customers operating indigitall every day, validating the platform reliability and AI capabilities behind OmnIA Agents.
430+ verified reviews across 8 categories, measuring delivery performance, AI capability and platform reliability in production.
Recognized for AI conversational capability across global, EMEA, and European markets.
The fastest-growing platform in the category: ranked above all competitors on traction and growth trajectory.
Highest usability score in the category: teams configure and run OmnIA Agents without heavy IT overhead.
Top-rated platform in Europe and Latin America by verified customer rating, in two separate G2 Regional Grids.
Most AI deployments have a hidden cost that doesn't appear in the contract: every customer query sent to a public large language model is a data transfer to that provider's infrastructure. indigitall AI Agents use Retrieval-Augmented Generation trained exclusively on client-provided content. The agent knows your products, policies, and procedures. It knows nothing it wasn't taught. And everything it was taught stays inside your environment.
Trained on your content: product catalogs, FAQs, policy documents, service data. Answers stay limited to that validated content.
Customer queries do not leave your environment to be processed by a third-party LLM provider. The reasoning layer runs on indigitall's infrastructure with your data.
Every response is filtered for brand safety, legal compliance, and content moderation before delivery. Blacklists prevent specific terms, competitors, or off-limits topics from appearing in any response.
More about AI Voice
indigitall AI Voice places outbound calls from the Customer Journey, triggered by customer behavior or prediction, and its documentation also describes inbound lines. During the call the AI agent reads from and writes to connected systems, and afterwards the journey can send an SMS, WhatsApp message, email, RCS or push, or hand the contact to a person with the context.
Call modes
Outbound campaigns and inbound lines.
Scale
10 to 10,000+ simultaneous calls, per the AI Voice documentation.
Systems
Connected through MCP or API; confirm the specific connection with indigitall in the sandbox test.
After the call
The journey can send SMS, WhatsApp, email, RCS or push, or transfer to a person.
Best for: customer engagement teams that need a journey to trigger a call, read live information from their systems, take an action during the conversation, record the result and continue by message or with a person when needed.
This page is for customer engagement, CX and operations leaders deciding whether an AI voice agent belongs in their journeys. It shows which calls the agent places, what it reads and writes during the call, which controls it has and which questions to settle before launch.
The journey decides who to call, when and why. Each use case below starts from an event, and the agent can continue on another channel without hanging up.
Trigger: an e-commerce cart is abandoned and cheaper channels such as WhatsApp and push have not worked. The agent calls the shopper to resolve doubts. Next step: a payment link on another channel, or a transfer to a person.
Trigger: a payment fails. The agent calls and the journey can follow with a payment link on another channel, such as WhatsApp, SMS or email.
Trigger: an appointment is not confirmed, as in a healthcare clinic. The agent reads booking information, checks availability and can confirm, reschedule or cancel the appointment, then sends a confirmation message before the call ends. The industry cases in the documentation include calling patients 48 hours ahead to confirm attendance; confirm the lead time and the confirmation channel for your deployment.
Trigger: a policy is about to expire, a delivery changes or an alert needs a response. The agent makes a courtesy call and can continue on the channel the customer prefers.
Choose AI Voice
Choose AI Voice when the contact needs a conversation and a message is not enough: a high-ticket cart, a failed payment, an appointment to confirm. Many people screen calls from numbers they do not know: in a Pew Research Center survey from July 2020, 81% of US adults said they do not generally answer their cellphone when an unknown number calls. That is why the order of channels matters in a journey: cheaper channels go first and voice follows when the margin justifies it.
Do not choose it
Do not choose it when a message already resolves the case, when you need a human contact-center operation (AI Voice hands calls off to a person with context, so confirm which operation receives them), or when you cannot show that the contact agreed to receive calls. Escalate to a person when the agent cannot resolve the question, and define in the sandbox which other situations, such as an urgent request or a complaint, should also reach a person.
Start with one trigger and one use case, list the systems the agent must read and write, and test the call in the sandbox with your own scenarios.
| Stage | What happens | Examples |
|---|---|---|
| Before the call | The customer profile gives the journey the context for the call. | Who the customer is and what happened last time. |
| Reads during the call | The agent queries live systems while the contact is on the line. | CRM record and calendar availability. Other systems depend on what is connected and are confirmed in the sandbox. |
| Writes during the call | The agent updates connected systems. | Books, reschedules or cancels an appointment, unsubscribes a customer on request, sends a one-time password. Further updates depend on the connected systems. |
| After the call | The result is recorded and the system returns status events to the journey, which decides the next step. | For example call_completed, payment_confirmed and transferred_to_human, as the integration documentation lists them. |
| Follow-up | The journey continues on another channel or with a person. A call nobody answers can route to an SMS or email follow-up. | SMS or RCS when there is no WhatsApp, push if the customer has the app, a payment link. |
AI Voice acts on your live systems through MCP, an open standard for connecting AI applications to external systems, so the agent can read before and during the call and write the outcome while the contact is still on the line. The integration method for each system is confirmed in the sandbox. The Customer Data Platform provides the profile, and the journey decides what runs next.
To place an outbound call, the journey uses a webhook node that sends an HTTP POST to the AI Voice service with the contact's properties, such as phone number, first name, order ID and custom context. The system returns status events to indigitall, for example call_completed, payment_confirmed and transferred_to_human, and the journey decides the next step. The documentation describes four implementation stages: design, test in a sandbox, scale and optimize. The exact systems, fields and data each use case reads and writes should be confirmed with indigitall in the sandbox test before launch.
Handoff to a person
When the agent cannot resolve a question, it can transfer the call to a human agent with a real-time transcript and a summary of the context, so the customer does not start from scratch. Which other situations, such as an urgent request or a complaint, should also reach a person is defined for your campaign and confirmed in the sandbox.
Consent and quiet hours
Consent and timing follow the AI Voice integration documentation. It states that you can only call customers who have agreed to receive outbound calls, that a customer who withdraws consent is flagged automatically and will not be called again, and that the agent tells the customer at the start of each call that they are speaking with an AI agent. It also recommends quiet hours, for example no outbound calls between 9 PM and 8 AM in the recipient's local time zone. These are product controls and documentation guidance, not a statement of legal compliance for any specific campaign.
| Criterion | What it should answer | indigitall AI Voice | Confirm before launch |
|---|---|---|---|
| Trigger | Which event or prediction starts the call. | A journey event or prediction, such as an abandoned cart or a failed payment. | Which of your systems can send that event to the journey. |
| Call model | Outbound or inbound. | Both. | Which mode your first use case needs. |
| Scale | Documented simultaneous capacity. | 10 to 10,000+ simultaneous calls. | The peak concurrency your campaigns need. |
| Action | What the agent can read and change. | Reads and updates connected systems; the connection method is confirmed in the sandbox. | Which systems and fields are connected for voice today. |
| Follow-up | Which channel runs next. | SMS, WhatsApp, email, RCS, push or a transfer; a call nobody answers can route to an SMS or email follow-up. | The fallback you want when nobody answers. |
| Voicemail | What happens if voicemail answers. | The public integration guide describes routing the user to a secondary SMS or email follow-up when a call goes unanswered or reaches voicemail. | The exact behavior in your journey, whether the agent leaves a message and what it says. |
| Transfer | When a person takes over, and with what context. | A transfer to a human agent with a real-time transcript and context summary when the agent cannot resolve the question. | Which contact center or phone system receives the call, and how urgent requests or complaints are routed. |
| Language | What is documented about switching languages. | Multi-language support with regional accents and local slang; the language is set at the start of each call, including from a customer attribute, and different phone numbers can be used for different languages and countries. | The languages and accents your market needs, and whether the language can change during a call. |
| Contact control | Consent and opt-out, without a legal claim. | The AI Voice integration documentation states that you can only call customers who agreed to receive outbound calls, and that a customer who withdraws consent is flagged automatically and will not be called again. | Your own consent records and your legal team's review of each campaign. |
| Calling window | When calls should not be placed. | The documentation recommends quiet hours, for example no outbound calls between 9 PM and 8 AM in the recipient's local time zone. | The hours your legal team requires and how you apply them in the journey. |
Yes, on the systems that are connected. indigitall AI Voice acts on connected systems while the contact is on the line: it can read the record and the calendar, and schedule or reschedule an appointment. The integration documentation also lists unsubscribing a customer on request and sending a one-time password. The exact systems, fields and write-back behavior are confirmed in the sandbox before launch.
Yes. The Customer Journey triggers the call from an event or prediction, and AI Voice also supports inbound lines.
If the call is unanswered or reaches voicemail, the public integration guide describes routing the user to a secondary SMS or email follow-up. Confirm the exact behavior in the sandbox, including whether the agent leaves a message and what it says.
The AI Voice documentation recommends setting quiet hours in the journey that triggers the call, for example no outbound calls between 9 PM and 8 AM in the recipient's local time zone. The hours a specific campaign needs are a decision for your legal team.
The AI Voice integration documentation states that you can only call customers who have agreed to receive outbound calls, and that a customer who withdraws consent is automatically flagged and will not be called again. At the start of each call the agent tells the customer that they are speaking with an AI agent. These are product controls and documentation guidance, not a statement of legal compliance for any specific campaign.
The AI Voice documentation says that when the agent cannot resolve the customer's issue or question, it can transfer the call to a human agent automatically. The person receives a real-time transcript and a summary of the context. The exact conditions for your campaign, including whether an urgent request or a complaint should also reach a person, should be confirmed with indigitall in the sandbox test before launch.
indigitall AI Voice transfers a live call to a human agent with a real-time transcript and context. The public AI Voice pages do not list which telephony or contact center systems receive that transfer, so confirm the one you use with indigitall in the sandbox test before launch.
The indigitall AI Voice agent writes the outcome to the connected systems it updates, and the system returns status events such as call_completed, payment_confirmed and transferred_to_human to the journey. When a call is transferred, the person receives a real-time transcript and a summary of the context. The AI Voice documentation also describes post-call sentiment data and resolution-rate tracking. The public AI Voice pages do not state how long call data is retained, so confirm retention and audit needs with indigitall before launch.
The AI Voice documentation describes multi-language support, with context and local slang handled across languages and native regional accents. The integration guide says you decide which language the agent uses at the start of each call and can set it from a customer attribute; it also mentions local variants and different phone numbers for different languages and countries. Confirm the languages and accents you need with indigitall in the sandbox test.
According to indigitall's AI Voice page, no. The agent uses retrieval-augmented generation trained only on content the client provides, customer queries do not leave your environment to be processed by a third-party language model provider, and the reasoning layer runs on indigitall's infrastructure.