MCE Systems
Agentic Voice AI for Telecom Device Support
Problem Statement
Telecom support is often reactive and disconnected, with customers only reaching out when something goes wrong. This leads to high demand, inconsistent experiences and frustration when trying to get help.
Operators needed a more connected approach that could support customers earlier, reduce issues and create a smoother overall experience.
Goals & Objectives
Transition from reactive to predictive and proactive device care
Deliver a real-time conversational AI agent for customer support
Create a smooth, connected experience across app, chat, phone and store
Unlock contextual upsell and engagement opportunities
Research & Insights
Support is mostly reactive, with little focus on preventing issues early
Device data creates an opportunity to predict and reduce common problems
Experiences are fragmented, forcing customers to repeat steps across channels
Customers expect fast, simple and conversational support
Approach
I designed the experience around how people naturally speak, creating a voice-first interaction rather than a traditional screen-based interface.
The interaction focuses on timing, interruptions and clear spoken responses, so conversations feel natural, easy to follow, and responsive in real time. The system also keeps track of what has already been said, so users don’t need to repeat themselves during the journey.
To support a realistic prototype experience, I created a bespoke voice using ElevenLabs and embedded it into the Figma prototype, helping establish a consistent and natural conversational tone for the AI agent.
The interface was kept intentionally minimal, allowing voice to lead the experience. Visual elements act as support rather than competition, with a simple text layer available for users who prefer to follow along on screen.
User Testing
The prototype was tested in real-time with colleagues and customers interacting with the voice AI agent to explore usability, clarity and overall experience.
The experience leverages Amazon Bedrock and AWS Nova Sonic, along with NVIDIA Nemotron, to power real-time conversational intelligence and natural spoken responses within the voice agent.
Final Designs
Final UI designs for the voice AI agent, showing the clean, minimal interface that supports conversational interaction and provides optional visual guidance.
Showcased at MWC & GTC 2026
The AI agent and prototype were presented at MWC 2026 and GTC 2026, where it was discussed with industry leaders and partners. Attendees engaged with live demos and had the opportunity to interact with the experience directly.
