Hyundai Motor : Scaling AI Voice Agents from Brazil to 3 Countries

Outsourced call center connection rate: 8%. For every 100 leads called, 92 never picked up. Hyundai Motor Global brought in BringTalk's AI Voice Agent to change that number — and what started as a single-country pilot in Brazil passed enterprise security audits in Australia and expanded into warranty and recall scenarios in the United States.

The Challenge — One Bottleneck Shared Across 3 Countries

Hyundai Motor Global's lead response process had a structural bottleneck. Meta ad campaigns drove test drive leads, and outsourced call centers followed up. The problem was simple — they were too slow.

• Call connection rate: 8% — 92 out of 100 leads never even connected

• Lead-to-Sales conversion: ~1% — 99% of ad spend effectively evaporated

• Response time: hours to days — the golden window had already closed

This wasn't a single-market problem. Brazil, Australia, the US — the bottleneck was the same everywhere. Only the regulations and use cases were different.

Phase 1: Brazil — Proving It with Numbers

The first production deployment was Hyundai Motor Brasil. An LQA (Lead Quality Assessment) agent was deployed to make outbound calls to Meta ad leads within 2 minutes of sign-up. The results were unambiguous.

Metrics kept climbing through iterative optimization. Valid Call Rate improved from 23% to 34.4% (+49%), Lead-to-Opportunity from 9.5% to 21.4% (×2.25). The contract was renewed.

Phase 2: Australia — Building on Top of Compliance

Brazil's numbers opened the door to HMCA (Hyundai Motor Company Australia). But Australia's regulatory landscape was different.

1. Australian Privacy Act — legal requirements for the entire data collection and processing lifecycle

2. Privacy Impact Assessment (PIA) — enterprise-grade security audit. Fail any item, no production

3. Australian Telemarketing Standard 2017 — telecommunications regulations governing outbound calls

BringTalk passed the PIA with a Zero Retention architecture — zero PII residue on any external server. Agent "Dai" greets leads in Australian English ("G'day"), calling Meta campaign leads for the i30 Sedan Hybrid and ELEXIO test drives instantly.

Dealer assignment is fully automated. A Follow-up Agent checks in 10 days later to verify customer satisfaction. Real-time Slack-based call monitoring catches anomalies instantly. The key: this wasn't a copy-paste. Brazil's architecture was adapted, not cloned — redesigned for Australia's regulations, language, and culture.

Phase 3: United States — Beyond Sales

At Hyundai Motor America, the scenario itself changed. Not sales — After-Sales. Warranty and recall. Agent "Sarah" covers two scenarios.

Recall Outbound

Proactive calls to IONIQ 5 owners affected by the ICCU safety recall. Explains the issue, confirms free repair, and connects to the nearest dealer — all in one call. Target call duration: 35 seconds.

Warranty Inbound

When a customer calls in, the agent looks up their VIN from a license plate, checks current mileage against warranty limits, and delivers real-time coverage status. Bumper-to-bumper, powertrain, EV battery, anti-perforation, emissions — all five warranty types assessed instantly, with dealer transfer if needed.

Backend Response Times (n8n automation)
───────────────────────────────────────
Customer lookup     < 50ms
Warranty check      < 100ms
Dealer transfer     < 150ms

Key Insight — Prove, Validate, Expand

This case proves one thing. Once an AI Voice Agent delivers results in one country, the next country is a compliance exercise. And the one after that is a scenario expansion.

Three countries. Three languages (Portuguese, Australian English, American English). Three regulatory frameworks (LGPD, Australian Privacy Act, US state laws). One Zero Retention architecture running through all of them.

It started as "sales automation." It ended as "automating the entire customer experience."