Ukraine's State Employment Service has deployed a generative AI voice assistant in its contact center through a pilot programme with ElevenLabs, marking the first use of such technology for incoming call handling at a Ukrainian state body. The assistant has processed 2,700 calls during the pilot phase, reducing average response time to 18 seconds and achieving near-complete coverage during after-hours and air alerts—periods when human agents are unavailable or overwhelmed. During standard business hours, the system handles approximately 20% of routine inquiries, freeing agents for more complex cases. The Ministry of Economy plans to scale from the current 4,000 to 80,000 monthly calls, suggesting confidence in the model's viability at higher volumes.
The deployment raises critical questions about how public sector contact centers will compete with private enterprise implementations. Where government agencies can demonstrate cost-effective, 24/7 availability through voice AI, the pressure on commercial CX teams to justify human-only models intensifies—particularly for high-volume, routine interactions. This is especially relevant for teams managing similar call patterns: if a state employment service can achieve 18-second response times and near-total after-hours coverage, what operational justification remains for contact centers still relying on traditional IVR or delayed callback systems? The 20% deflection rate during peak hours also signals that voice AI isn't positioned as a replacement for agents but as a triage layer, which aligns with how enterprise platforms like Genesys and Salesforce are positioning agentic capabilities.
The choice of ElevenLabs as the technology partner—a specialist in multilingual voice synthesis rather than a traditional contact center vendor—suggests a shift in how contact center capabilities are being assembled. Rather than monolithic platform upgrades, organizations are integrating best-of-breed voice AI into existing infrastructure. For CX leaders evaluating their tech stack, this raises a practical consideration: should you be building voice capabilities through your primary platform vendor, or are specialized voice AI providers now mature enough to justify point-solution integration? The Ukrainian case demonstrates the latter is operationally viable at scale.
A voice AI assistant has been launched in the contact center of the State Employment Service dev.ua