If, in the last two years, the emergence of Generative Artificial Intelligence has served as a glimpse of the potential of this technology – and, we must agree – had a reasonable impact in areas such as customer service, in 2025 we should witness the development of "agentic AIs", that promise to substantially transform the landscape of technology. Alongside the ever-increasing expansion of AI models to an even wider range of companies and niches, the fact is that, today, no company can ignore the potential application of AI in innovation or operations.
Unlike traditional AIs, that require constant human supervision, agent-based AIs are designed to operate independently, performing complex tasks without direct human intervention. This advancement is made possible by deep learning algorithms that allow systems to understand and process large volumes of data in real time, quickly adapting to new information and contexts
Furthermore, agent-based AI systems use large amounts of data from various sources to analyze challenges independently, develop strategies and execute complex and sequential tasks. The potential application of this type of AI is enormous, starting with customer service, going through the processing of any type of information or processes of the company, and also for cybersecurity, where it is possible to automate tasks that today require human intervention, how to analyze and fix vulnerabilities in systems, for example.
In Brazil, the adoption of agentic AI is still in the early stages. Some sectors are already testing the new model, and according to a survey conducted by the Institute of Applied Economic Research (IPEA), by 2025, about 40% of large Brazilian companies plan to integrate agent-based AI systems into their operations
Impact of agentic AI
The potential impact of agentic AI is enormous. Banks and financial institutions could reduce fraud incidence by up to 50% with technology, according to the Brazilian Federation of Banks (FEBRABAN).
The health sector will also be able to apply the new technology. The Brazilian Medical Association (AMB) highlights that agentic AI has the potential to reduce medical errors by up to 30%, since technology is capable of analyzing medical records, test results and health history of patients to propose more accurate diagnoses. In the industry, intelligent automation will be driven by agentic AI, that allows the operation of machines and processes autonomously.
Expansion of generative AI to the productive environment
Even with the spread of the use of generative AI, its impact has still been low in the productive environment, with more intense use in some niches, as image and video creation. According to Gartner, the adoption of this AI model is expected to increase in the productive environment by 2026 – reaching adoption by up to 80% of companies.
In Brazil, the adoption of generative AI tools by companies is growing, as organizations recognize the value of these technologies in process optimization and innovation. Companies from various sectors, including advertising, media, and design, have been using generative AI to create personalized content and more effective campaigns.
Furthermore, large corporations are beginning to integrate generative AI into their daily operations to enhance data analysis, the automation of repetitive tasks and the forecasting of market trends. The adoption of these tools can transform the way Brazilian companies operate, increasing efficiency and competitiveness in the global market
AI will become increasingly humanized
The launch of ChatGPT-5 is expected to happen in the coming months, and one of the most anticipated features of this new version is the enhanced ability of the tool to maintain natural conversations. This means that the chatbot will be able to follow the flow of a conversation, understand the context and the hidden meaning, and even respond "emotionally"
Furthermore, specialists have suggested that GPT-5 will have reasoning abilities similar to those of humans, being able to understand the context of a conversation in a more comprehensive way
2025: the year of small AI models
When AI emerged, the learning models known as LLMs – Large Language Models have been massively adopted for popular tools to emerge in the market. These models are trained on large amounts of data – however, this information is more superficial.
Small models are cheaper to build and operate and are more easily adapted to specialized applications. Instead of trying to do everything, small models are customized to perform a more limited set of everyday tasks for a specific business need
LLMs have billions of parameters and require massive amounts of data and computational power to train and run. Small models, on the other hand, can be effectively trained with less data and require much less computational power, therefore, energy) to execute
In summary, these changes promise to transform various sectors and bring significant innovations to the daily lives of people and businesses. The advancement of AI, both in terms of accessibility and sophistication, will further democratize access to advanced technologies, paving the way for a future in which technology will be deeply integrated into all aspects of society.
With the proliferation of smaller and more specialized AI models, it is expected that personalization and efficiency will reach new heights, providing solutions increasingly aligned with the specific needs of each sector. Therefore, the year 2025 promises to be, without a doubt, a year of great revolutions for AI.
