Delivering on that mission at scale is a different challenge entirely. Nova Gestões generates close to 100 million calls per month plus 45 million additional contacts via WhatsApp, SMS, and email, reaching 5 to 7 million unique clients monthly across a team of 1,700 employees. Quality teams could manually review just three to five calls per agent per month, leaving 99% of interactions invisible. "We had an invisible operation," Braz says. "I did not know the voice of the client, I did not know why they were not paying or what offer might solve their problem."

The manual process created problems everywhere. After each call, agents spent two minutes entering data into the CRM, a process riddled with errors and human bias. Quality monitors, working from tiny samples, introduced their own biases into performance reviews. Hiring was increasingly difficult, leaving a growing gap between operational demand and available talent.

When generative AI arrived as a market force, Nova Gestões didn’t chase the hype. They identified real pain points and attacked them one by one. After testing other providers and finding quality results came at costs 10 times higher than acceptable, Google Cloud and its Gemini models stood out for delivering the right balance of language comprehension, scale, and cost. Partner IPNET by Vivo supported the implementation from the start, ensuring the right technical foundation was in place from day one.

"We already arrive at the call knowing who that person is, what happened, and how to speak with them," Braz explains. "That changes everything."

The second solution, Borderless Service, uses Gemini to translate conversations in real time. Foreign professionals, many of them highly qualified individuals including engineers and doctors who couldn’t previously work in Brazil due to language barriers, now handle calls in their native language while clients receive fluid Portuguese.

Standard translation tools were tested and rejected because they couldn’t handle Brazilian slang, regional dialects, and the specific vocabulary of debt negotiation. Generative AI solved what libraries couldn’t.

For Braz, the broader shift in the collection industry is already underway. "The competitive difference will no longer be who makes the most calls," he says. "It will be who understands the client more precisely and reaches them in a more human way." AI will not shrink the workforce but reshape it, replacing volume with expertise and generalist operators with specialists who are better supported and better informed.

For other contact centers considering a similar move, his advice is direct: start with a real problem, build a solid cloud foundation before deploying any model, and measure everything. "Do not implement just because of the hype," Braz says. "Apply small solutions that attack real pain points and deliver results you can actually see."