Meet UVA, the AI Co-Pilot

We built an AI Co-Pilot for Customer Support Operating Teams

Gen AI

Telco

NOS, a telecommunications and media company, was facing challenges in their customer support operations. We helped them by simplifying access to internal company information, leading to shorter calls and faster resolution times.

Implementation time:

3-5 months

estimated annual savings

€118k

correct answers

87%

calls powered by the AI solution

20%

data related to the pilot phase

The Opportunity

The support team knew that customers needed fast and clear answers, but they were blocked by:

Search Difficulties

Operators struggled to find answers due to slow information retrieval mechanisms, causing long waiting times for customers

‍

Pages Blocked

Over 15,000 pages scattered across disconnected systems

‍

Inconsistent Services

Clients received inconsistent service, resulting in longer calls and extended hold times

The Solution

We built UVA, an AI virtual assistant. This AI tool receives questions from customer support operators, searching the company’s internal knowledge base at blazing speed and providing a timely, relevant answer to assist operators in responding to clients.

‍

While developing the agent, we also discovered that some of the existing information was outdated, duplicated or even inaccurate, which resulted in UVA producing incorrect answers. We addressed this by not only gathering the information, but self-validating it. UVA provides only verified answers, while also allowing operators to rate responses and update content when needed.

‍

As with most AI-driven tools, change management is a key topic we prioritized from the start, focusing on user experience and maintaining a tight feedback loop with the customer support team. Their involvement throughout the development phase helped ensure UVA was aligned with their needs. We focused on adoption, and this improved the AI output immensely.

Meet the team of this project

Hugo Veiga

Senior Machine Learning Engineer

Isabel Labarca

Data Scientist

The Impact

In the call centers, operators have the power to use Generative AI that complements the manual search process. This shift enables them to access information faster, improving both service quality for customers and satisfaction among the staff. As of August 2025, UVA is in the pilot phase, accessing all 15k pages and rolled out to a single team.

‍

UVA's responses are achieving 87% accuracy, based on evaluations by the operators themselves. Inaccurate answers are flagged and corrected at the source, creating a self-improving system.

‍

But more than the technical metrics, the company is estimated to save up to €118,000 per year, with UVA’s pilot phase alone. Our client is eager to expand UVA beyond the pilot phase, aiming to roll it out not only across all customer support teams but also across multiple areas of the company.”