How was the first "AI in Practice" event, promoted by AWS and BlueMetrics?
- Marcelo Firpo
- 4 days ago
- 13 min read
AI-generated summary:
The first “AI in Practice” event, hosted by AWS and BlueMetrics at the Caldeira Institute and streamed live on YouTube, showcased the concrete application of generative artificial intelligence and machine learning in the corporate environment. Aimed at guests from the innovation ecosystem, the event brought together BlueMetrics experts to demonstrate how companies can transform hype into real results with rapidly implemented, high-impact projects. Six use cases developed for clients in Brazil and abroad were shared, along with discussions on data governance, technological choices, the risks of public AI, and adoption strategies. The event highlighted BlueMetrics' position as a leader in data and AI solutions, with nearly 200 projects delivered in a decade of operation.
A morning of learning and exchanging experiences.
Despite the low temperatures on the morning of July 3rd, the Campus Auditorium of the Caldeira Institute in Porto Alegre was packed for the "AI in Practice" event, promoted by AWS in partnership with BlueMetrics. Aimed at guests from the corporate ecosystem and also broadcast live on YouTube, the meeting aimed to present practical and strategic applications of generative artificial intelligence (GenAI) and machine learning (ML) in the business world. As the name suggests, the focus was less on concepts and more on rapid execution and the generation of real results, with concrete examples of projects implemented in different sectors.
During the event, executives and experts from BlueMetrics demonstrated how robust data engineering, the use of AWS technologies, and agile development methods allow companies to implement AI solutions in just a few weeks, with a direct impact on operational efficiency, decision-making, and new digital experiences.
Six real-world case studies developed for clients in the education, media, financial, and real estate sectors were presented, all illustrating how AI can quickly become a tangible competitive differentiator for companies.
With ten years of experience, BlueMetrics has established itself as a leader in data analytics and, more recently, in generative AI projects. The company has already completed nearly 200 successful deliveries for approximately 100 clients in the United States, Brazil, and Latin America, combining technical excellence, pragmatism, and strong execution capabilities in complex data and artificial intelligence projects.
Far beyond the hype: AI that delivers real results.
Gabriel Casara, CGO of BlueMetrics, opened the “ AI in Practice” event by reinforcing the company's commitment to practical AI applications and its strong partnership with AWS. He presented market data that highlighted the consolidation of artificial intelligence as a transformative technology: according to him, 72% of global companies had already adopted some AI solution by 2024, and private investments reached US$ 252 billion.
Casara highlighted that AI has surpassed the "hype" status and has begun to generate real value, with projections of an impact of up to US$15.7 trillion on the global economy by 2030. According to him, the disruption goes beyond the operational level and is already affecting strategic decisions, such as scenario analysis, product development, and even the creation of new business models.
He also drew attention to the need to contextualize AI within the business environment. Casara said that while generic AI, used daily by individuals, is powerful, corporate solutions need to be adapted to the specific context of the business. He emphasized that the value of AI lies not only in the algorithms, but in how it is applied.
Casara also emphasized the critical importance of data, since "good data generates good AI ." The executive stated that many AI projects end up requiring a restructuring of the company's data, because, even with a large volume of information available, the data is not always organized or ready to generate value. He compared the use of AI to a Formula 1 car: extremely powerful, but which does not generate results if driven on a poorly paved dirt road. This road, according to Casara, is the data, which needs to be well prepared for the AI to function efficiently.
“ Price Waterhouse believes that we will have $15.7 trillion added to the economy by 2030 (with the AI revolution). The disruption is not just in operational tasks: it also affects the strategic issues of companies. ”
The perfect storm for AI adoption.
The following presentation, by Denis Pesa, CEO of BlueMetrics, began with a clear vision: artificial intelligence has ceased to be just a trend and has become a concrete tool for generating real value in business. Denis began his speech by presenting the company's new positioning, "data and AI solutions for the real world ," and reported that, in 2024, BlueMetrics took a strategic pause in its operations to thoroughly study the practical impacts of AI.
The goal was to reposition the company to deliver applicable solutions with a quick return, going far beyond the hype generated by the subject. According to him, the great challenge was precisely to transform this hype into measurable results, something the company had already been achieving in dozens of projects with clear gains in productivity and revenue.
Denis contextualized the current moment as a "perfect storm" for AI adoption: accessible cloud infrastructure (with AWS support), accelerated evolution of LLMs, data availability, and greater openness from companies to technological adoption. He presented examples of applications already in operation, such as intelligent virtual customer service, content generation, fraud detection, credit analysis, predictive maintenance, and personalization in e-commerce. An important highlight was the replacement of traditional dashboards with generative AI interfaces, which make data more accessible to managers and accelerate decision-making.
The CEO also warned about the risks of the indiscriminate use of public AI in corporate contexts, such as data leaks, violations of the LGPD (Brazilian General Data Protection Law), model hallucinations, and reputational damage. He emphasized that companies need to invest in customized, secure, and contextualized models, avoiding known cases of misuse that have generated real losses.
Denis concluded by emphasizing that AI should be treated as a strategic business agenda and encouraged companies to start with high-impact, low-cost MVPs, relying on the technical support offered by BlueMetrics and the financial support, in the form of MVP funding , offered through the partnership with AWS. He ended with an optimistic provocation: "The future is already here. Let's embark on this journey together."
“ The hype is now generating real results, financial results. That's a fact. We know this from the dozens, almost a hundred projects we've worked on in the last 18 months, implementing this technology for our clients. ”
The increasingly strategic role of AI agents.
In the third morning presentation, Fabiano Saffi, CTO of BlueMetrics, addressed the role of generative AI agents, which he described as " a hype within a hype ." He explained that agents are autonomous systems that act as an intermediary layer between the user and learning language models (LLMs), allowing these models to operate with greater assertiveness, security, and contextualization. Saffi highlighted that, unlike simple prompt -based interactions with generic AI, agents structure the decision-making process, integrate private company data, and generate more relevant and secure responses for the business.
Saffi explained that the autonomy of agents means they can plan actions, break down complex problems into steps, use memory, access structured databases, trigger external APIs, and even interact with machine learning models. All of this is done in an orchestrated, secure, and controlled manner, through limits and validations. One example was an agent for the restaurant sector, capable of providing information about dishes, reserving tables, and automatically updating the menu without human intervention.
He also highlighted a real-world application developed for an American client operating in the real estate investment trust (REIT) sector. In this project, BlueMetrics created multiple agents integrated with machine learning and generative AI models, capable of compiling financial data, analyzing scenarios, generating reports, and offering strategic insights to managers. According to Safi, this was possible thanks to the correct structuring of the agents, something that had not been achieved by other vendors who attempted solutions based solely on prompt engineering . In two and a half months, BlueMetrics delivered a robust solution, which is already in the final approval phase.
Finally, the CTO of BlueMetrics presented the emerging concepts of " Generative BI " and " Generative Analytics ," which enable the automated creation of dashboards and analyses via natural language. He emphasized that these technologies, combined with intelligent agents, have great potential to improve engagement with data and transform how companies analyze indicators, develop diagnoses, and make medium- and long-term decisions.
“ If we provide the model with data in a structured way, in a secure environment, with compliance regarding our data, the response will be more complete. And that's one of the main reasons why we all started talking about agents. ”
Three GenAI case studies, with deliveries made in record time.
The following presentation, featuring case studies, was given by Diórgenes Eugênio (Dio), Head of GenAI at BlueMetrics. He shared three practical cases of operational transformation using generative AI, highlighting the speed and impact of these solutions. Dio began by reinforcing the objective of making the application of GenAI to businesses tangible, bringing concrete ideas that could stimulate the audience's creativity. The first case presented was from an educational group with a high demand for customer service during enrollment periods. For this client, BlueMetrics developed an intelligent virtual assistant that, using GenAI and multiple agents, is able to answer questions in a contextualized, scalable, and efficient way. The system uses filters to redirect sensitive topics, integrates with private databases, and enriches context, all within an architecture built in just 10 weeks.
In the second case, Dio tackled a project with a major TV network focused on generating personalized content at scale. The client was struggling to transform news videos into textual content adapted for multiple platforms. The solution involved automatic transcription with AWS Transcribe, followed by unbiased summaries generated by GenAI. From this, specific texts were created for Instagram, Facebook, X/Twitter, and blogs, respecting tone, audience, and platform. The key differentiator was the use of multiple specialized templates for each channel, which increased the accuracy and personalization of the generated content. The entire pipeline was developed and delivered in six weeks.
Finally, Dio presented a third case focused on the digitization and contextualization of historical print newspapers. The client possessed archives from the 1980s at risk of deterioration. After digitizing the material with AWS Textract, BlueMetrics applied OCR and pre-processing techniques to differentiate headlines from body text. GenAI organized the articles in a structured way, and then all the content was indexed in a vector database. This allowed for intelligent contextual searches, going beyond keywords and offering a much richer experience. The solution was created in eight weeks, even with high complexity and a high degree of technical creativity.
In conclusion, Dio emphasized that these three projects—in education, media, and documentary archives—illustrate the power of generative AI to transform operations in weeks. He highlighted that the use of multiple agents with specific functions and accessible services, such as Amazon Bedrock, allows for the creation of robust solutions with speed and low cost, provided they are well-planned.
“ My goal here is that, by the end of this presentation, you will all be able to materialize some ideas, sharpen your creativity, and think about how you can bring this into your daily life, your business, your operation. ”
Want to see GenAI and Machine Learning solutions
making a difference in your company?
Next, three highly strategic case studies based on Machine Learning.
After the coffee break , it was time for the presentation by Bernardo Trevizan, Head of Data & Analytics at BlueMetrics. He discussed the role of Machine Learning within the artificial intelligence ecosystem, distinguishing it from generative AI.
He explained that ML is a subset of AI, focused on pattern detection and predictions based on historical data. Generative AI, such as LLMs (Large Language Models), is a product of ML focused on language, such as summarizing texts or generating content. Bernardo highlighted that many business challenges require models beyond language: for example, to predict default, detect fraud, or estimate future revenues.
The first case presented was the optimization of credit granting in the real estate sector. BlueMetrics' client was performing this process manually, which generated inconsistency and risks. A classification model via Amazon SageMaker was implemented, which assigned ratings based on data such as income, marital status, and number of children. The model achieved 92% accuracy in predicting good payers and also allowed for customer segmentation for marketing campaigns, prioritizing those with a good payment record. Based on historical data, it is possible to predict that this solution will reduce delinquency by 46%.
In the second case, Bernardo explained the development of a financial fraud detection model via Pix (Brazil's instant payment system), for a banking software company. Without prior data, an unsupervised ML model was chosen, capable of learning the standard behavior of individual and corporate accounts. When a new transaction was analyzed, the system checked if it was statistically outside the norm, and, if so, generated an alert of possible fraud. Using real-time inference in SageMaker, the model began delivering responses in less than 1 second, with the potential to block up to R$ 1.5 million in fraud.
The third example was an integrated solution developed for the US real estate capital markets, combining data engineering, generative AI, and machine learning. The system allowed managers to request revenue forecasts for specific properties via natural language. The AI recognized the limitations of LLMs in predicting the future and triggered an agent with time series models in SageMaker, which returned estimates with a 5% margin of error.
According to Bernardo, this solution replaced managers' intuition with unbiased analyses, accessible through a conversational interface. He concluded by emphasizing that machine learning is a continuous process that requires constant reassessment and training to maintain its accuracy over time.
“ Through machine learning, we were able to analyze customer profiles (good and bad payers). And we achieved 92% accuracy. In other words, every time the solution told me that a rating was A, I was 92% certain that it was, in fact, an A. ”
The BlueMetrics way of delivering with agility and efficiency.
In the sales area presentation, Gabriel Casara (CGO) and Luciano Rocha (CCO) presented BlueMetrics' strategic approach to data and artificial intelligence projects. Casara began by explaining that many companies arrive with the desire to implement AI, but still lack clarity about the real problem. BlueMetrics' role is precisely to help clients identify whether the challenge lies in the data structure, the application of machine learning, or the use of generative AI, and to act in an end-to-end manner , with the technical capacity to deliver from engineering to the final application. He emphasized that the company has a strong partnership with AWS, which provides the secure, scalable, and flexible foundation for the solutions, allowing BlueMetrics to be creative and agile even in complex projects.
Luciano Rocha highlighted the importance of companies treating the use of AI as a strategic issue, not an improvised one. According to him, new generations of employees are already using AI naturally in their daily work, which increases the responsibility of leaders to structure these technologies within the organization safely and purposefully. He explained how BlueMetrics structured its business verticals, focusing on machine learning, generative AI, and data foundation, which has been the heart of the company since its inception. Luciano emphasized that behind every successful AI solution lies a well-structured database.
Casara resumed speaking to present blue4AI , BlueMetrics' proprietary method based on four simple and practical steps:
Discovery and design
Proof of concept (PoC)
Implementation (Deployment)
Continuous optimization
He explained that this model allows for rapid results in 6 to 10 weeks, with deliverables in the form of production-ready MVPs. The company has already applied this method to over 190 projects with a high degree of success, including for international clients.
Casara concluded by reinforcing the company's competitive advantage in offering fast deliveries, end-to-end solutions, and high flexibility, whether by allocating entire teams or developing projects on demand, all with a focus on generating real and continuous value for clients.
“ Our role here is to help clients look inward, to structure their data, to do so securely, and to accelerate these projects in a way that truly makes sense. ”
“ We usually say, "There's no such thing as 'no,' right? Maybe there's 'not like this,' because we'll solve it in some other way. We have a lot of experience today, with over 190 implemented AI data projects and more than 95 satisfied clients. "
Questions and answers and final considerations
The question and answer session complemented the practical journey presented in the lectures. Diórgenes Eugênio (Dio) answered the first question, about how BlueMetrics chooses the best solution or technology for each project. He explained that it all starts with clarity about the business problem, and that the team segments the challenges into three areas: data, machine learning, or generative AI.
Within each vertical, a technical evaluation is conducted that takes into account cost, quality, and latency, always focusing on the sustainability and scalability of the solution. Bernardo Trevizan added that the team's role as a technical consultant is reinforced, capable of translating high expectations into real and effective solutions, adapted to the data and infrastructure maturity of each client.
The second question came from an industry representative who wanted to know how to apply machine learning in environments with data scattered across different systems (ERPs, CRMs, etc.). Bernardo responded by highlighting the importance of data engineering and the creation of a data lake , where all data is centralized and organized to support AI models.
He explained that different algorithms are used depending on the nature of the problem (classification, regression, clustering) and that the process involves both science and experimentation. Furthermore, he emphasized that it's possible to adapt knowledge from one sector to another. For example, reusing credit rating models to categorize leads or emails. For industry clients, we recommend reading the case study we developed for a large truck and bus manufacturer, in which we were able to reduce process analysis time from 4 hours to 6 seconds.
In the third question, regarding the use of data lakes , Bernardo and Fabiano Saffi explained that the choice depends on each objective: a data lake is ideal for exploring raw and varied data, while a data warehouse is geared towards structured analysis, generating KPIs and insights.
In his closing remarks, Denis Pesa, CEO of BlueMetrics, shared a personal reflection on his journey. He recounted how for many years he felt frustrated because he saw value in technologies that were not yet recognized as essential in Brazil.
Now, with the popularization of generative AI and its concrete application in the local market, Denis celebrated the moment as a true turning point: "first-world cutting-edge technology is finally within our reach . "
Denis also highlighted that, upon seeing the presentation of some of BlueMetrics' most recent case studies, he realized that the company is not only replicating existing solutions, but, through its extensive technical expertise and multidisciplinary talent, is actually creating new and different solutions to better address the specific challenges of its clients.
He concluded by thanking the team, AWS partners, and participants, and extended an invitation for companies not to let ideas stagnate and to begin transforming their challenges into real projects with technical, strategic, and financial support.
" Seeing these presentations we had here today, it's clear that we're managing to do different things, really new things. "
This was our summary of the first AI in Practice event, presented by AWS and BlueMetrics. The full presentation can be accessed here.
Did you like the case studies presented? Are you interested in applying technologies like GenAI and Machine Learning to solve your company's specific challenges? Let's talk about it.
Learn about some Use Cases .
Comments