“We went from having data to actually using it”: Make helps Bemol deliver a small-business care on enterprise scale

Bemol, a leading Brazilian retail company with 5,500+ employees, has a dedicated collections team that follows up with customers who fall behind on installment payments. They collect thousands of call recordings and text replies where customers share their problems. These conversations are full of real customer signals and opportunities to improve the customer experience. 

But with a volume of data that high, much of it never got to the people at Bemol who could act on it. Customer complaints often disappeared in the sea of data. Warranty issues stayed hidden from the responsible team. Leading to loyal customers churning.

Nelson Manfre, Senior Collection Analyst on Bemol's Collections Strategy Team, helped change that. Using Make, together with Whisper AI and ChatGPT, he built a set of two AI agents. 

The first one listens to collection calls and routes anything actionable, like a customer mentioning a product defect, straight to the team that should handle it. The second AI agent reads every reply to Bemol's automated payment texts and flags the ones that need a human response. Together, these agents transformed idle data sources into a live feedback loop that reduced churn. 

Ever since, Nelson has repeated this approach when he sees a problem: he builds an automation himself, proves it works, and embeds it in how the team works. The engineering team is no longer required for his team to solve problems; they build by themselves. 

"The big-picture win? We went from having data to actually using it, and that shift is generating value in ways we hadn't even planned for when we started."

The challenge: Vital insights lost in a sea of data

Bemol records every call their collection team makes. Customers share useful feedback during those calls. Before Make, none of that information went anywhere useful. There was no way to keep up, let alone route it to a team that could actually follow up.

"If a customer says they didn't pay because a product had a defect, that information needs to reach the warranty team, so they can follow up with that customer individually. Before this, that kind of personalized handling simply didn't happen at scale."

Bemol also sends automated text reminders to customers who are behind on payments, and plenty of customers reply. Every one of those replies got saved to a database, but with that much volume, the team never had the analytical capacity to read through them and figure out which ones needed a reply.

The solution: A small-business care at scale

Nelson’s first Make orchestrated AI agent transcribes collection calls using Whisper AI and analyzes them with ChatGPT. 

The agent identifies the specific, actionable information that customers mention (like a defective product or a warranty issue) and automatically routes it to the team that needs to act on it. 

A complaint that used to disappear the moment the call ended now reaches the warranty team on its own, allowing individual follow-ups at the rapid scale.

A flagged text reply got the right support to an elderly customer 

The second agent reads every customer reply to Bemol's payment texts and prioritizes the ones that actually need attention, instead of leaving them to sit in a database. 

One of those flagged replies came from an elderly customer in an isolated community. He'd texted back that his TV had stopped working. Thanks to the AI agent, that reply reached the right team, who found out the TV was still under warranty and arranged a replacement.

"We would never have seen this message if we hadn't been looking. This customer lived somewhere isolated, he wasn't very familiar with technology, and he had no way to come into a store himself."

The results: “We solve a problem, customers return”

Personalized, individual follow-up on issues like warranty complaints now happens automatically, at a scale that wasn't possible before. Text replies that used to go unnoticed are read, prioritized, and acted on, turning idle data into live insights. 

None of it required a meaningful increase in manual effort; the same team is simply no longer missing key information that was already coming through.

This level of customer service is crucial for a company that operates across an Amazon region. People from isolated towns, like the elderly customer, can fully trust Bemol’s online store; they will stay loyal and spread a positive word-of-mouth.

"When we solve the problem of one client, he is happy, and he will return to our store. This is what would never happen if we didn't have this agent."

Building customer trust, one AI agent at a time

Make was Nelson's first real contact with AI agents. Building with Make helped him learn how AI agents work in practice.

"I come from a business background. Make lets me build AI agent workflows in a simple but powerful way. That experience deepened my understanding of prompt engineering and agent orchestration, and I think that's exactly where Make shines: putting advanced capabilities into the hands of business people."

Nelson's AI agents have since grown from a small pilot to over 100,000 calls a month, earning direct backing from Bemol's CEO and management team. The business value of these AI agents has proved so strong that Bemol now screens for AI literacy in hiring, at a time when many Brazilian companies remain cautious about the technology.

What's most rewarding for Nelson, though, is the customers who keep returning because of the trust that is built. Like the elderly man whose voice almost went unheard. AI agents that make it possible to listen to customers across huge volumes of data ensured he still has a working TV.

This is the service Bemol keeps building toward with every AI agent. 

A service people can trust.