Automated phone calls API: trigger AI calls from your CRM
Learn how to connect an automated phone calls API to your CRM, web forms, and internal databases to run AI outbound calls and sync outcomes back automatically.
## What you can automate with an automated phone calls API
An **automated phone calls API** connects your software to voice agents that place outbound calls and return structured data. When a trigger event occurs, your CRM, website, or backend system sends a payload that launches an outbound call based on your configured rules.
DeepAgent provides programmatic triggers to automate core operational workflows:
* **Call a new web form lead:** Reach out while prospective buyers are still browsing your landing page.
* **Call from a contact list:** Launch batches of contacts without manual dialing.
* **Call on CRM lifecycle changes:** Dial automatically when a lead reaches a specific stage, owner, or status property.
* **Call before an event:** Deliver reminders and confirm attendance for upcoming workshops, consultations, or bookings.
* **Handle inbound customer care:** Answer incoming lines around the clock, triage technical issues, and log support tickets directly in your helpdesk.
Once each call finishes, the platform sends back structured outcomes: call disposition codes, audio recordings, transcripts, booked calendar slots, and logged support tickets.
## Which events can start a call
You can trigger calls based on specific customer actions or operational states inside your company:
| Event | Source system | What the agent does | Real example |
| :--- | :--- | :--- | :--- |
| Web form submitted | Website / Landing page | Calls within seconds, qualifies investment interest, books consultation | Flor de Pacifico case study |
| New lead generated | WhatsApp / Lead flow | Reaches out on WhatsApp first, then places a phone call using collected data | Bludental case study |
| Event registration | Form / Event calendar | Calls attendees, shares logistics, confirms attendance, logs records | Lunaga case study |
| Delivery failure case | Logistics back office | Calls recipient 7 days a week to collect redelivery or return preferences | Italian logistics operator case study |
| Inbound lead batch | CRM marketing list | Screens industrial fit and hands warm prospects to sales reps | Italian manufacturing company case study |
| Inbound support call | Incoming phone line | Triages technical problems, verifies customer ID, creates service tickets | Italian water-treatment company case study |
In real estate investment, Flor de Pacifico configured their setup so calls start within seconds of web form submissions. Their voice agent completed 207 calls across 5.5 days, securing 76 live answers and 18 qualified appointments, which amounted to a 25% conversion on answered calls between 08:00 and 21:00. This prompt connection model relies on the principles outlined in our guide to speed to lead.
Bludental, an Italian dental clinic chain with 70+ locations managing 10,000 monthly leads, uses a sequential multi-channel trigger. First contact happens via WhatsApp within 5 seconds of lead generation. If needed, a phone call follows after about 1 minute using the initial qualification details collected in the chat. You can study this multi-channel architecture in our guide on combining WhatsApp and voice AI.
For event operations, Lunaga connected an agent to their registration form and calendar for an online workshop. Across 18 days, the system handled 10,133 outbound calls, achieving 770 confirmed attendees while logging each call record to a Google Sheet without human intervention.
Logistics teams also run programmatic workflows. An Italian integrated logistics operator uses an outbound agent named GIULIA. When parcels enter a failed delivery state, the back-office system dispatches case data via API. GIULIA calls recipients across an 08:00 to 18:00 window, completing 18,364 calls over 20 weeks with 1,549 positive delivery instructions recorded.
For inbound workflows, an Italian water-treatment company uses an agent named SARA. SARA handled 2,260 inbound calls in one month, answered inquiries around the clock, achieved 1,730 resolved cases, and opened tickets automatically in the CRM.
## API trigger or list upload: when each fits
Depending on team structure and technical requirements, you can trigger calls either event-by-event via API or in bulk via file uploads.
| Attribute | Event API trigger | List upload campaign |
| :--- | :--- | :--- |
| **Speed** | Instant execution per lead | Scheduled batch dispatch |
| **Setup effort** | CRM workflow or webhook configuration | Exporting and uploading a CSV file |
| **Best for** | Inbound leads, urgent notices, transactional updates | Outbound sales lists, seasonal promos, re-engagement |
A **webhook voice agent** integration works well when speed matters. You connect your landing page to execute automated outbound calls from web forms immediately. The customer submits a form, the webhook fires, and the phone rings.
List uploads suit cyclical outbound campaigns. In the DeepAgent platform, you upload your CSV file, map the phone and variable columns, assign your agent persona, and start the campaign. An Italian manufacturing company used this batch approach to process 2,000 inbound B2B leads in cohorts of 500 contacts, cutting manual handling time by over 70%. When considering cold prospect lists, review the parameters discussed in our breakdown of AI cold calling.
## The data a call needs
To run a contextual conversation, your systems pass key properties when initiating a call. Sending rich data lets the agent greet contacts by name, reference previous touchpoints, and ask the right questions.
Essential data properties include:
* **Contact name:** First name and last name for proper greeting.
* **Phone number:** The telephone number formatted in international standard format with country code.
* **Consent and source:** Proof of opted-in consent and lead capture origin.
* **Call reason:** Context tag explaining why the call is happening.
* **Context fields:** Specific data points (product requested, clinic branch, order number) used in the script prompt.
* **Language preference:** The language model to use during the interaction.
* **Callback rules:** Permitted hours and maximum retry attempts.
* **External ID:** The record identifier from your CRM to match writeback data accurately.
Here is an illustrative example of an outbound call payload with placeholders:
```json
{
"lead_id": "",
"first_name": "",
"last_name": "",
"phone_number": "",
"preferred_language": "",
"reason_for_call": "",
"custom_context": {
"source_form": "",
"product_interest": "",
"service_location": ""
},
"consent_confirmed": true
}
```
*Note: This code block is an illustrative example with placeholders, not a technical API specification for the DeepAgent platform.*
## Attempts, hours and stop rules as configuration
Setting clear boundaries on call frequency protects brand reputation and respects customer schedules. An **AI calling API** lets you establish calling windows, retry limits, and stop triggers.
As a configuration example from their case, Bludental applies a recall sequence set to `+0h / +6h / +12h / +18h` after the initial contact, while enforcing a rule to never place calls after 19:00. This configuration contributed to a conversion rate of 28% for the AI agent compared to 22% for human agents, alongside an answer rate of 65% versus 50% for manual teams.
In another configuration example from our logistics operator case, agent GIULIA was set to place up to 6 attempts per day between 08:00 and 18:00, running 7 days a week. Across 20 weeks, 64.2% of contacts were reached on the first call, while 35.8% answered only on subsequent retries.
The Italian industrial equipment manufacturer configured an operating window of 08:30 to 17:30, Monday to Friday, capped at a maximum of 5 attempts per lead. For detailed frameworks on cadence management, read our complete guide to lead follow-up automation.
Automated queues also require stop rules. The system drops the lead from further attempts if the contact requests an opt-out, if the line disconnects as invalid, or if the customer indicates they are not interested.
## What comes back after the call
A **voice AI API** returns structured data once a conversation ends. DeepAgent sends conversational metadata back to your software through a webhook.
Standard outcome categories returned include:
* **Reached and goal met:** The prospect answered, qualified, and scheduled an appointment or completed the verification.
* **Callback requested:** The contact asked for a follow-up at a different time.
* **Not interested:** The prospect declined the offer.
* **No answer / Busy:** The line rang without pickup or hit a busy tone.
* **Wrong number:** The person answering is not the named contact.
* **Opt-out / Do-not-call:** The contact demanded removal from contact lists.
Together with the outcome code, DeepAgent returns the complete transcript, an audio recording link, any chosen meeting timestamp, and field extractions such as support ticket notes. In the Lunaga campaign, outcomes flowed directly into a Google Sheet. For the industrial manufacturing company, outcomes logged 100% automatically into HubSpot deals and contact records. To connect these pipelines, consult our guides on HubSpot and Salesforce voice integration and our detailed DeepAgent HubSpot setup guide. Detailed performance tracking practices are also discussed in our article on AI call analytics.
## Checks before going live
Before activating automated calls on live records, walk through this operational checklist:
1. **Consent verification:** Confirm that you store documented consent for every lead record before initiating outreach.
2. **Duplicate filtering:** Prevent duplicate calls by establishing deduplication checks based on email, customer ID, or phone number.
3. **Phone formatting:** Validate that all numbers follow international telephone standards with valid country prefixes.
4. **Internal test dials:** Run end-to-end test scenarios on company test numbers to review voice pacing, pronunciation, and webhook delivery.
5. **Webhook error handling:** Build fallback mechanisms in your endpoint to capture retries if your CRM server drops a connection.
6. **Human transfer routing:** If your agent offers live escalation, confirm destination phone lines or call center queues are staffed during campaign hours.
7. **AI disclosure:** Comply with European transparency rules, including EU AI Act Article 50, by having the agent introduce itself as an artificial intelligence assistant at the start of the call.
8. **DPO review:** Review calling hours, legal bases, and national do-not-call registries with your DPO or legal advisor. Review our compliance manual on GDPR and AI voice agents, along with our analysis on the limits and opportunities of automated sales calls.
## DeepAgent pricing
DeepAgent provides transparent subscription tiers to accommodate teams from launch to enterprise scale:
* **Starter:** €99 per month
* **Pro:** €299 per month
* **Growth:** €500 per month
* **Scale:** €799 per month
* **Enterprise:** Custom quote based on organizational needs
All quoted prices exclude VAT. Choosing an annual billing plan offers a 17% savings. You can explore full platform features with our self-service trial offering 30 minutes free, no card. Visit our pricing page for more information.
## Frequently asked questions
### Can I start an AI phone call from my CRM?
Yes. When a record enters a pipeline or changes status, a workflow fires an HTTP request to the voice AI API. The agent calls the lead using variables from that CRM record and writes the outcome back.
### How do I send website form leads to a voice agent?
You can connect your form builder directly through a webhook voice agent trigger or pass the submission through your CRM. The agent receives the prospect contact details and places the call according to your configured rules.
### Can the outcome go back into the CRM automatically?
Yes. When the call ends, DeepAgent pushes the call status, transcript, audio link, and custom fields like meeting dates or support tickets back into your CRM properties.
### Can I upload a list instead of using the API?
Yes. If you run one-off outbound campaigns, you can upload a CSV spreadsheet directly in the DeepAgent platform, map your fields, pick your voice agent, and launch the campaign.
### Do I need a developer to set this up?
No developer is needed for native connections like HubSpot or manual CSV uploads. A developer or marketing operations specialist is helpful if you build custom API payloads or private database webhooks.
## Trigger your first automated voice call
Connecting an automated phone calls API turns static database records into immediate conversations without adding manual dialing to your team's schedule.
## Related reading
If you want to dig deeper, these DeepAgent articles cover related topics:
- Hotel AI Receptionist: 24/7 Phone and WhatsApp Handling
- Are AI phone calls legal? TCPA, PECR, and consent
Talk to sales → · Create your agent →