SouJava Sessions was created with a clear purpose: to bring even more value to the
community by creating a space for presentations, workshops, and talks that, for different
reasons, may not fit into the monthly in-person meetups.

From the beginning, the message was direct: this is a space built by the community
for the community. It is not only about Java, and it is not only about formal talks.
It is about creating an open platform for people who want to share knowledge, exchange
experiences, and help others grow in their careers.

And to open this journey, there could hardly be a better guest than someone who brings
together experience, didactics, community presence, and a practical vision of the future
of technology.

AI is not a sudden revolution, it is an evolution

One of the first points raised by Professor Isidro was how we look at artificial intelligence.
Much has been said about the AI revolution, but he proposed a more accurate reading:
what we are living through is an evolution.

Artificial intelligence did not appear overnight. What we see today is the result of
decades of research, experimentation, mistakes, progress, and technological maturity.
Tools such as ChatGPT, Copilot, Cursor, Claude, Gemini, and many other models are the
result of a path that began long before the recent popularization of generative AI.

This perspective matters because it removes some of the exaggeration around the topic
and brings the discussion closer to the reality of developers. The question stops being
only “will AI replace programmers?” and becomes:
how can we use AI to eliminate repetitive tasks, create better products, and increase delivery capacity?

The end of low-value repetitive work

One of the most memorable concepts from the talk was the idea of using AI to eliminate
repetitive, operational, low-value work. In a humorous tone, Professor Isidro referred
to this kind of task as “corno job”: boring, recurring, exhausting work that consumes
the time of people who could be solving more important problems.

The core message was simple: AI should help remove from the routine what does not require
much creativity, analysis, or sophisticated human decision-making. Resetting passwords,
classifying requests, routing support tickets, summarizing documents, searching internal
knowledge bases, and executing repetitive workflows are examples of activities that can
be automated or assisted by AI.

When this happens, the developer’s role changes. They stop being only someone who writes
code for traditional CRUDs and become someone capable of building solutions that add more
value to the business.

The real challenge is not just using AI. It is knowing where AI truly creates value
and how to integrate it safely into the systems that already exist.

Chatbots, assistants, and agents: each one has its place

Another important point of the session was the difference between chatbots, assistants,
and agents. Although many people use these terms as if they were the same thing, they
represent different levels of autonomy and responsibility.

A chatbot works with us. It responds, suggests, and helps, but still
depends heavily on human direction. Code copilots fit very well into this category.

An assistant works for us. It can summarize documents, organize information,
generate materials, analyze content, and support more complex tasks.

An agent works on our behalf. It interprets a situation, makes a decision
within defined limits, accesses tools, queries databases, executes actions, and can interact
with other systems.

This is where the conversation becomes more serious. If an agent can act, it also needs
clear boundaries. It is not enough to connect AI to a database, a CRM, or an internal API
without thinking about security, context, permissions, and responsibility.

Guardrails: the necessary boundaries for intelligent agents

The talk also reinforced the importance of guardrails. In AI applications,
guardrails define how far an agent can or cannot go.

This care is essential because agents can query data, create records, open support tickets,
trigger workflows, respond to customers, and interact with real systems. Without well-defined
boundaries, a solution that seemed productive can become an operational, financial, or security risk.

For those who develop enterprise software, this point is fundamental. AI cannot be treated
as magic. It needs to be architected, monitored, validated, and integrated with the same care
we apply to any critical system.

When the MVP becomes a real product, Java steps in

During the session, Professor Isidro also shared an interesting reflection on no-code and
low-code tools such as N8N, Make, and Zapier. They are excellent for prototyping, validating
ideas, and allowing non-technical people to automate processes.

But when an automation grows, starts serving more users, needs to scale, be monitored, have
observability, cost control, logs, security, versioning, and integration with corporate systems,
a new question appears:
how do we turn this MVP into real software?

This is where Java becomes a protagonist.

The argument was not an empty defense of the language based on personal preference. The idea
was to show that Java remains one of the main technologies in the enterprise world. Large banks,
payment systems, streaming platforms, e-commerce platforms, and critical applications use Java
because they need robustness, predictability, scalability, and long-term maintenance.

In other words: if AI needs to enter real corporate systems, it makes sense for Java to be part
of that conversation.

Spring AI, LangChain4j, and the Java ecosystem for AI

The talk also covered frameworks and libraries that bring generative AI closer to the Java world,
especially Spring AI and LangChain4j.

Spring AI appears as a very natural alternative for those already working in
the Spring ecosystem. It reduces the learning curve and makes it easier to integrate with LLM
providers, embeddings, vector databases, and Spring Boot-based applications.

LangChain4j, on the other hand, was presented as a more flexible option for
working with different models, creating chains, customizing conversational memory, integrating
tools, building agents, and connecting Java applications to different AI providers.

One of the most interesting points was the possibility of creating observability around AI model
consumption. In real applications, it is not enough to know whether the answer is correct. It is
also necessary to understand token usage, cost per user, cost per conversation, peak usage periods,
and the financial impact of the solution.

This vision brings AI closer to real software engineering. It stops being just a beautiful chat
interface and becomes part of an architecture that needs to be measured, maintained, and evolved.

RAG, knowledge bases, and practical applications

Another concept covered was the use of supporting knowledge bases, especially in
RAG scenarios. The idea is to allow an application to use documents, texts,
PDFs, internal databases, or specific content as an additional source of knowledge to answer
users more effectively.

Instead of relying only on the general knowledge of a model, the application can combine LLMs
with business data. This makes it possible to create assistants capable of answering questions
about products, policies, internal documents, study materials, company processes, or any other
relevant content.

Professor Isidro also demonstrated practical examples, including a Java bot integrated with
Discord, capable of using a knowledge base to guide students on projects and even perform code
review from a Java file.

This kind of example clearly shows the potential of AI when combined with a mature language,
good architecture, and a real problem to solve.

The Java developer in the age of AI

The main message of the session was that Java developers do not need to stay on the sidelines
of the AI evolution. On the contrary: those who already work with Java have a strong foundation
to build intelligent solutions within corporate environments.

Instead of abandoning the current stack and moving to another technology just because AI became
popular in other ecosystems, Java teams can leverage their existing skills to build prototypes,
create APIs, integrate AI providers, work with vector databases, implement agents, apply guardrails,
and bring artificial intelligence into real systems.

This is an important vision for companies and professionals. AI does not need to be a complete
break from what already exists. It can be a natural evolution of the architectures that already
support critical applications in the market.

A first edition worthy of the community

As the first edition of SouJava Sessions, the meeting with Professor Isidro fulfilled its role
very well: it opened space for a relevant, current, and deeply practical conversation connected
to the real challenges of those who build software.

The session was about technology, but also about career, community, responsibility, and the future.
It showed that Java remains alive, modern, and ready to actively participate in this new phase of
artificial intelligence.

It also reinforced the purpose of SouJava Brasília: to create spaces where people can share knowledge,
gain visibility, ask questions, learn together, and strengthen the Java community in Brazil’s Federal
District and beyond.

SouJava Sessions began with a clear message: the community has a lot to share, and this space is open
to anyone who wants to contribute.

And if the first edition already brought a conversation at this level, it became clear that much more
was yet to come.

SouJava Brasília continues as a living movement: where knowledge, technology, and community
come together to shape better professionals and build more relevant solutions.