Direct, citation-backed answers to the questions European operators ask most often.
Is AI voice agent deployment compliant with the EU AI Act?
Yes. Voice agents are classified as Limited Risk under the EU AI Act, triggering the transparency obligations of Article 52: the caller must be informed they are interacting with AI. We script this disclosure into the opening turn of every conversation and log the disclosure event. Full enforcement applies from August 2, 2026, with penalties up to 15 million EUR or 3% of worldwide turnover for transparency violations (Article 99).
Where is call audio stored?
On EU-hosted infrastructure only. Default regions are AWS Frankfurt (eu-central-1), AWS Ireland (eu-west-1), Azure Sweden Central, OVHcloud Gravelines and Scaleway Paris. No call audio, transcript or embedding is transferred outside the European Economic Area at any point. This is aligned with Schrems II (CJEU C-311/18), CNIL February 2026 guidance on Microsoft 365, and the Austrian DSB March 2026 ruling on US data transfers.
What latency do callers actually experience?
End-to-end response latency averages under 1 second on our production deployments. This is achieved through three engineering choices: native audio-to-audio models (no STT then LLM then TTS hop), an EU-hosted inference endpoint geographically close to the telephony provider, and a streaming output pipeline that begins emitting audio before the model has fully generated the response.
How many languages are supported?
Six native-accent models in production today: English, Spanish, Dutch, French, Arabic and German. The system detects the caller's language in the first turn and adapts mid-call when the caller switches. Additional languages can be added in the deployment phase based on your traffic profile.
Can voice agents handle GDPR Article 9 special category data?
Yes, with the appropriate lawful basis. Healthcare and insurance deployments often involve special category data. We configure explicit-consent capture per call, encrypted at-rest storage with EU-controlled keys, and Data Protection Impact Assessment under GDPR Article 35 produced during the deployment phase.
How does pricing work?
We operate on a deployment plus monthly retainer model, not per-minute reselling. You receive a fixed deployment cost (typically billed across four weeks) and a monthly operational retainer that covers infrastructure, model licensing, monitoring, compliance updates and quarterly reviews. Contact us for a tailored proposal.
What happens when the AI cannot answer?
Human escalation is triggered by three default conditions: explicit caller request, low confidence on intent detection, or detected emotional distress via real-time sentiment analysis. The full conversation context (transcript, intent history, sentiment trace) transfers to the human agent so the caller never has to repeat themselves. This satisfies EU AI Act Article 14 on human oversight.
Can I use AI voice agents for outbound calling under EU law?
Yes, with compliance controls. Outbound campaigns require a documented lawful basis under GDPR Article 6 (typically consent or legitimate interest), respect for the ePrivacy Directive on unsolicited communications, Do Not Call list enforcement (CTPS in UK, equivalent in each Member State), and time-of-day restrictions per local law. We build all of these into the outbound campaign manager.
Do AI voice agents work with my existing CRM?
Yes. Native integrations exist for HubSpot EU, Salesforce EU Hyperforce and Pipedrive EU. Custom CRMs integrate via REST API and webhook. Every call produces a structured record with caller identifier, transcript, intent, outcome, sentiment trace and next-action recommendation, posted to your CRM in real time.