Initial Situation
Starface, a leading provider of Unified Communications solutions, identified the strategic need to transform traditional telephony with modern AI technologies. Conventional IVR (Interactive Voice Response) systems limit the user experience through rigid menu paths ("Press 1") and a lack of flexibility.
The goal was to develop an AI Voice Agent, that not only answers calls but also conducts natural dialogues, intelligently pre-qualifies inquiries, and dynamically adapts to the context. The solution also had to meet the highest requirements for latency, data security, and scalability to function as an integral part of the Starface platform.
Our Solution
AMAI developed a high-performance voice agent architecture in close collaboration with Starface, deeply integrated into the cloud infrastructure as a Python microservice. The solution combines the flexibility of generative AI with the reliability of traditional telecommunications.
High Configurability & Multi-tenancy:
Unlike "one-size-fits-all" solutions, the system offers a high degree of customization. Starface customers can individually configure the agent's behavior – from tonality to specific routing rules. This enables seamless adaptation to different corporate identities and service processes.
Massive Scalability & Performance:
The architecture is designed for high-load scenarios. The service dynamically scales to several thousand parallel streams, without compromising performance. This ensures constant availability and consistent service quality even during high call volumes (e.g., during peak hours).
Enterprise-Grade Security & Guardrails:
Using LLMs in customer interactions requires strict security mechanisms. We implemented robust Guardrails, which ensure the agent stays within its role and does not disclose protected data (PII) or hallucinate. Quality assurance is achieved through automated test scenarios that prevent breakout attempts and guarantee compliance with regulations.
Real-time Audio Pipeline & SLA Compliance:
For natural conversation, latency is the critical factor. Our optimized audio pipeline (Speech-to-Text, LLM inference, Text-to-Speech) processes speech in real-time, ensuring minimal delays that comply with strict Service Level Agreements (SLAs). The result is a fluid dialogue that feels human.
Seamless System Integration:
The Voice Agent is not an external component but is natively embedded in the Starface architecture. Via gRPCinterfaces and modern state management, the Python service communicates directly with the telephone system. As a containerized Kubernetesapplication, it integrates seamlessly into existing CI/CD pipelines and the deployment concept.
Results & Business Impact
The introduction of AI Voice Agents marks a technological leap for Starface. The solution transforms telephony from a static input channel into a dynamic, intelligent service touchpoint.
- Increased Efficiency: Automated pre-qualification and information gathering significantly relieve human agents.
- Competitive Advantage: Starface offers its partners an innovative, future-proof solution that clearly stands out from the market standard.
- User Experience: End customers experience natural interaction instead of frustrating menu navigation, which increases acceptance and customer satisfaction.
Customer Feedback
"Our collaboration with AMAI was particularly strong because they were extremely pragmatic and solution-oriented. Especially in the AI field, you quickly realize during a project that architectural decisions are rarely a one-and-done deal 😄
New challenges, dependencies, and technical details constantly emerge that weren't on our radar initially. And that's precisely where the collaboration worked incredibly well.
The team quickly immersed themselves in our architecture, processes, and requirements to develop genuine solutions alongside us.
It was particularly important for us not to view AI in isolation, but always to consider aspects like scalability, operations, integration, and long-term flexibility.
Ultimately, this resulted not only in a technically robust AI Voice Agent but, more importantly, a collaborative partnership where we learned a great deal together."
Jennifer Rapp, Starface GmbH, Project Management AI Services
.svg)
