AI and agents
Hyper-personalized AI voice calls from your customer data platform: how AI360 connects to voice agents
AI360 turns the customer profile into personalized outbound voice calls. It decides who to call, why, when and in which language, checks consent, and sends each customer's approved details as variables to a voice agent platform such as ElevenLabs Agents or the calling platform you already use. Call outcomes, answers and opt-outs flow back into the profile, so every next call is better informed.
Published . Last updated
On this page
Voice is still the channel customers answer for things that matter: a payment due, a renewal, a service problem. AI voice agents now hold natural conversations at scale. What they lack is the customer: who should be called, about what, with which details, and whether they agreed to it. That is the job of the customer data platform. This guide shows how AI360 connects the profile to voice agent platforms.
What is the difference between personalized and hyper-personalized calls?
| Level | What changes per customer | Example |
|---|---|---|
| Mass | Nothing | "Your payment is due soon." |
| Personalized | Fields in a fixed script | "Hi Sarah, your payment of $420 is due on Friday." |
| Hyper-personalized | The reason, timing, offer, language and conversation path | A call timed after payday, in the customer's preferred language, mentioning the app she uses, offering to move the date because her balance is low, and able to answer her questions |
Hyper-personalization needs the full profile and live signals, which is why it starts in the CDP, not in the dialler.
How does the integration work?
- Select the audience. AI360 Studio picks customers from scores and triggers: payment due, renewal near, churn risk, abandoned application, website intent.
- Check permission. DataNature Atlas checks consent for the purpose and channel, do-not-call lists, contact limits and permitted calling hours for the customer's location.
- Decide the content. Next-best action chooses the reason, offer and language; the best time to call comes from past answer rates.
- Build the variables. Each customer gets a small, approved payload: first name, language, product, amount, due date, offer code. Nothing else leaves the platform.
- Hand over to the voice agent. AI360 sends the list, or individual call requests for real-time triggers, to the voice platform.
- Capture the outcome. Results, structured answers and transcripts return by webhook or API, land in the lakehouse and update the profile.
How does AI360 work with ElevenLabs Agents?
ElevenLabs Agents run conversational voice agents over phone numbers connected through Twilio or SIP trunking, so you can keep your existing carrier and numbers. Agents speak more than 70 languages, including English, French and Arabic. For outbound campaigns, batch calling accepts a recipient list where each extra column becomes a dynamic variable in the conversation, and special columns can override the agent's language, first message, system prompt or voice per call. Calls can be scheduled, monitored and created through an API.
With AI360:
- AI360 creates batch calls through the API, one row per customer, with the approved variables as columns.
- For real-time triggers, such as an abandoned application, AI360 places individual calls through the API.
- The agent's tools can call a narrow DataNature endpoint, for example to check today's balance or the next advisor slot, scoped to that one customer.
- Post-call webhooks send transcripts, analysis and extracted data back to DataNature, where they are stored, masked and linked to the profile.
ElevenLabs offers regional data residency and zero-retention options on enterprise agreements; check which apply to your country before you launch.
Can we use another voice platform?
Yes. The pattern works with any voice agent or contact-centre platform that accepts a call list or API request and returns results: AI360 selects and personalizes, the voice platform calls, and outcomes come back. Many organizations use the platform their contact centre already runs, or a regional platform with stronger support for a local language or dialect. See the regional guides for Canada and the Middle East.
What comes back into the CDP?
| Returned data | Where it goes | How it is used |
|---|---|---|
| Call status (answered, no answer, failed) | Contact history in Silver | Retry logic and best-time-to-call models |
| Structured answers (promise to pay, rating, preferred date) | Profile and feature store | Next actions and segmentation |
| Transcript and summary | DocuContext partitions, with personal data masked | Quality review, topic analysis, Voir answers |
| Opt-outs | Consent store | Immediately block further calls |
| Outcome after the call (payment, renewal, upgrade) | Gold outcomes | Lift measurement against a control group |
How do you keep it compliant and trusted?
- Consent first. In the United States, the FCC's 2024 ruling confirmed that AI-generated voices count as "artificial" under the TCPA, so these calls need prior express consent, caller identification and, for telemarketing, an opt-out. Canada's Unsolicited Telecommunications Rules require express consent for automated telemarketing calls. Check the rules in each country you call.
- Disclose that the call is automated and who is calling.
- Never ask for secrets. Agents must not ask for PINs, one-time codes or full card numbers.
- Limit what the agent knows to the variables for that call, and keep sensitive fields out of prompts.
- Respect hours and frequency by time zone and local rules.
- Keep records: every audience, payload and outcome is logged in Atlas.
See AI outbound calling compliance for detail.
How do you measure it?
Hold out a control group for every campaign and compare outcomes: payments made on time, renewals, upgrades, completed applications. Track answer rate, opt-out rate and complaints alongside conversion, so personalization improves results without wearing out goodwill.
Frequently asked questions
Does customer data leave our country?
Only the minimal variables for each call go to the voice platform, and you choose the platform and its region. The profile, transcripts and outcomes stay in your DataNature deployment.
Can the voice agent answer questions about the customer's account?
Within limits you set: the agent can call a narrow, audited endpoint for that customer only, and anything sensitive is handed to a person.
Which use cases work best first?
Reminders with a clear outcome: payment and renewal reminders, document or verification renewals, and appointment confirmations. They are easy to measure and low risk.
Sources
- ElevenLabs: batch calling
- ElevenLabs: post-call webhooks
- ElevenLabs: SIP trunking
- Orrick: what the FCC's AI voice ruling means
How does DataNature Atlas keep this secure and compliant?
Agents are governed like users: scoped, masked, logged and unable to act without approval. DataNature Atlas is the platform's governance module:
- Knows your data: a catalog with automated PII detection and classification across Bronze, Silver and Gold.
- Controls access: role-based grants plus attribute-based policies, row filters and column masks (null, redact, partial, hash, year only).
- Governs AI: Voir agents are scoped principals; masks, row filters and PII guardrails apply to every tool call and answer.
- Proves it: lineage from source to metric to decision, and an audit log of who or what accessed which data.
- Stays local: Atlas runs inside your deployment, so policies, lineage and logs never leave your jurisdiction.
These controls support GDPR, PIPEDA, Quebec Law 25, CCPA, HIPAA and sector rules such as OSFI E-23. Read DataNature Atlas in depth or see security and compliance.
Platform components in this article
Learn more about each DataNature product and module mentioned here.
Products
See it on your own data.
Book a demoDemos and quotes are handled by Dlytica, the company behind DataNature.