Redesigning Voice Customer Support with AI

How we built a conversational virtual voice assistant for NOS to reduce call volume and improve customer experience

Gen AI

Media & Telecom

NOS aimed to evolve its traditional IVR based customer support into a more natural and efficient conversational experience. We developed a conversational AI Voice Assistant focused initially on billing and payments, integrating intent detection, contextual resolution, and controlled experience transitions. The solution combines conversational AI flows with existing static IVR guides, enabling higher automation while preserving user control and service quality.

Implementation time:

3 - 5 months

calls per month expected

-5k

estimated monthly operational savings

12k

maximum conversational turns per billing interaction

The Opportunity

Rigid and limited IVR experience

‍The existing voice journey relied on static IVR menus and reactive guides, allowing a maximum of two useful conversational turns, which limited resolution capacity and often led to customer frustration or unnecessary human escalation.

High volume of billing and payment contacts‍

Billing and payment use cases represented a significant share of inbound calls, creating operational pressure and increasing costs, despite many interactions being repetitive and suitable for automation.

Need for a hybrid conversational model‍

NOS required a solution that could introduce conversational AI without disrupting the entire support architecture, ensuring coexistence between new AI driven flows and legacy static flows while maintaining strict control over loops, fallback logic, and experience quality.

The Solution

We designed and implemented a conversational AI Voice Assistant integrated into NOS’s IVR ecosystem, introducing an intelligent intent detection layer capable of routing customers to either static or conversational flows. The initial MVP focused on billing and payment use cases, where a dedicated conversational agent interprets open questions, performs contextual disambiguation when needed, and delivers natural language resolutions using data from the myNOS application APIs. The architecture enforces strict control of conversational turns to avoid loops, includes fallback mechanisms to IVR menus when limits are reached or frustration is detected, and ensures a fluid, less robotized interaction. By combining structured IVR logic with AI powered conversational resolution, we enabled NOS to incrementally transition toward a more scalable and human like customer support model, while preserving operational stability and measurable business impact.

Meet the team of this project

Pedro Fonseca

AI Engineer

The Impact

The implementation of the conversational Voice Assistant enables NOS to significantly reduce operational pressure on its customer support teams while improving the overall customer experience. By automating high volume billing and payment interactions, the solution reduces unnecessary transfers to human agents and shortens resolution times for common queries.

The projected reduction of 10,000 calls per month translates into meaningful cost savings and improved efficiency, while customers benefit from faster, more natural interactions. This hybrid conversational model also establishes a scalable foundation for expanding AI driven automation to additional use cases, supporting NOS’s long term strategy of modernizing and optimizing its customer support operations.