The most interesting generative AI success stories in Brazil don't come from research labs. They come from day-to-day operations: product registration, ad publication, IT support requests, and citizen services. These are processes with high volume and repetition, where generative AI ceases to be an experiment and starts appearing in the indicators.
In this article, we've compiled four projects that Flexa Cloud built on AWS with OLX, Gimba, Britânia, and FEBRABAN . For each, we show the challenge, what was done, and the measured results. Finally, we explain what these cases have in common and how to apply the same lessons to your company.
Summary: Results of generative AI in 4 companies
| Company | Challenge | Results with generative AI |
|---|---|---|
| OLX | Slow and incomplete ad posting. | Publication until 3x faster, more than 5,5 million ads |
| Gimba | Manual product registration in the catalog | Time per item 84% lower, from 13 to 2 minutes |
| Britannia | Overwhelmed internal service desk | 30% a 40% of the so-called daily tasks resolved by the assistant Tânia |
| FEBRABAN | To guide millions of people on how to deal with debt. | Chatbot trained with 15 years of data My Pocket Up-to-Date |
OLX: ads published up to 3 times faster
At OLX, the bottleneck was at the top of the funnel. Anyone wanting to sell something had to fill out several fields in the ad, and many people gave up or posted with incomplete information. This worsened the buyer's experience and reduced the quality of the platform.
The solution was a generative AI tool that automatically fills in the ad based on information already provided by the seller. The result: up to 3 times faster posting speeds , approximately 250 people using the tool every day, over 5,5 million ads created with it, and up to 10% more fields filled in some categories.
The solution is already used daily by around 250 people and has enabled more than 5,5 million advertisements.
Marcelo Wendling, CTO of the OLX Group
The full article about the OLX project was published on TI Inside.
Gimba: Product registration 84% faster with Amazon Bedrock
At Gimba, a distributor of office supplies and B2B e-commerce, the problem was product registration. Each new item required manual descriptions, attributes, and standardization, which delayed product entry into the catalog and generated inconsistent information.
With generative AI on Amazon Bedrock , the catalog is now automatically generated and reviewed by the team. The time per item dropped by 84% , from 13 to 2 minutes . Catalog data became more consistent, which also improved organic search ranking, with a projected 10% increase in search volume.
Gimba completely transformed its product registration process, reducing the time per item by 84%, from 13 minutes to just 2 minutes.
Daniel Arruda, CTO of Gimba
The details are in the Gimba case study and the case study published by AWS.
Britannia: Virtual assistant resolves up to 40% of IT calls.
At Britânia, the internal service desk received a high volume of simple calls, such as questions about access, procedures, and guidance. The IT team spent a good part of the day on repetitive tasks and had little time for problems that actually required a specialist.
The answer was Tânia , a virtual assistant with generative AI built on AWS and integrated directly into the ticketing system. Today, she answers between 30% and 40% of daily tickets , guides employees, and only passes on to the human team what needs analysis. The project came along with the migration to the cloud, which brought scalability, cost reduction, and governance.
The solution already accounts for approximately 30% to 40% of daily calls, automating simple support tasks, correctly guiding employees, and integrating directly with our call management system.
Paulo Beber, Senior Software Architect and AI Specialist at Britânia
FEBRABAN: Meu Bolso em Dia (My Pocket in Order) provides guidance to indebted individuals on a large scale.
" Meu Bolso em Dia ," FEBRABAN's financial education program, faced a scaling challenge: guiding millions of Brazilians on debt and default without compromising the accuracy of the information.
The program gained a chatbot with generative AI from AWS, built on 15 years of accumulated data. The user converses in simple language and receives guidance to resolve their own situation, with consistent answers based on the program's official content.
The chatbot, built on 15 years of program data, helps guide people in resolving debt and delinquency issues through a simple and intuitive interface.
Raul Moreira, chairman of Banco Original and board member of FEBRABAN.
What do these generative AI success stories have in common?
The four projects are from different sectors, but they follow the same pattern. There are five lessons that are valuable for any company:
- They began with a process involving volume and repetition. Registration, advertising, calls, and customer service. That's where generative AI delivers quick and easily measurable gains.
- They defined the metric before starting. Minutes per item, percentage of calls resolved, publication speed. Without indicators, there's no way to prove return on investment.
- They took care of the data. The quality of the response depends on the quality of the foundation. In Meu Bolso em Dia, we've had 15 years of organized content.
- They integrated them into the systems that already existed. Tânia interacts with the call system, and Gimba's registration feeds the catalog. AI has become part of the process, not a separate tool.
- They had governance from the start. Security, data access control, and human review where necessary. This is what separates a pilot program from a successful outcome. Understand the risks of... AI without governance.
How to bring generative AI to production in your company.
To replicate these results, most companies run into the same problem: they lack a team that understands the business, masters the AI architecture on AWS, and can prioritize the right initiatives. That's why Flexa Cloud's Center of Excellence (CoE) in Hyperautomation and Governed AI exists.
The Center of Excellence (CoE) organizes all of the company's AI ideas into a single queue, prioritizes them based on expected return, builds the solutions on AWS, and measures the impact of each delivery. In practice, it's the same method used in the cases above, with a senior team acting as an extension of your operation.
- Why are companies adopting the AI Center of Excellence?
- AI agents: how this technology is transforming operations
- More cases: Indra in the free energy market e Bravante in offshore decommissioning.
Frequently asked questions about generative AI in companies
What are some examples of successful generative AI case studies in Brazil?
OLX, which publishes ads up to 3 times faster with autocomplete; Gimba, which reduced product registration time by 84% with Amazon Bedrock; Britânia, whose virtual assistant resolves 30% to 40% of IT support requests; and FEBRABAN, with the chatbot for the Meu Bolso em Dia program.
In which processes does generative AI deliver results faster?
In processes with high volume and repetition, such as product registration, content creation, internal customer service, and customer support, the time savings are immediate and easy to measure.
How do you measure the return on investment of a generative AI project?
Defining the indicator before the project begins, such as time per task, volume processed, percentage of service requests resolved, or cost per operation, and comparing it to the previous situation after implementation.
How do I start a generative AI project on AWS?
Choose a high-impact process, define the success metric, prepare the data, and build with governance from the start. An AI Center of Excellence, like Flexa Cloud's CoE, helps prioritize initiatives and bring them to production.
Do you want results like those of OLX and Gimba?
Schedule a free diagnostic with the Flexa Cloud CoE team. We'll map out where generative AI can generate real gains in your operation and how to bring it to production with governance on AWS.
