CDL Goiânia already automates the management of 35,000 companies with AI agents in member support and churn risk monitoring. Simultaneously, Anthropic launched agents for the financial market capable of assembling presentations, reviewing financial statements, and escalating cases for compliance. The central point for businesses has shifted from the mere existence of technology to another: where it fits into operations with criteria, context, and supervision.
This shift is important because AI agents do not operate like fixed-flow chatbots. According to Serasa Experian, they are autonomous systems capable of perceiving the environment, making decisions, and acting to achieve specific objectives. And, according to TI INSIDE Online, the practical difference lies in their ability to understand context, learn, and adapt throughout execution.
What Differentiates AI Agents from Traditional Automations and Chatbots
The most relevant change is not in the name of the technology, but in the type of task it can perform.
Based on the facts gathered from the sources, AI agents combine three capabilities:
- perceive information from the environment
- make decisions to achieve an objective
- execute actions with independence or minimal human intervention
In practice, this distinguishes them from models based solely on pre-programmed responses. TI INSIDE Online emphasizes that agents need to understand the context to function effectively, precisely because they are not limited to fixed flows.
This point is clearly illustrated in the CDL Goiânia case. According to the report published in 24Cast, member support “no longer uses fixed-flow chatbots: there is a real knowledge base answering questions and escalating to humans when necessary.” The operation also includes an agent that monitors the behavior of the base and issues churn risk alerts.
The statement attributed to the organization summarizes the shift in stage:
“We don't treat AI as an experiment — it's an operation. Member support no longer uses fixed-flow chatbots: there is a real knowledge base answering questions and escalating to humans when necessary.” CDL Goiânia Representative
This type of application demonstrates an objective distinction:
| Technology | Operational Basis | Operational Limit Described in Sources |
|---|---|---|
| Fixed-flow Chatbot | Pre-defined rules and paths | Responds within closed scripts |
| AI Agent | Context, decision, and objective-oriented action | Responds, adapts, escalates, and executes tasks with more autonomy |
The accessibility of the technology itself is also increasing. Microsoft has published content guiding how to create and use the first AI agents, signaling that the topic has already entered the radar of corporate and technology teams.
Where Agents Are Already Being Applied in Businesses
The most concrete cases from the research material appear in three areas: customer service, finance, and legal.
1. Customer Service and Relationship Management
At CDL Goiânia, agents were applied to automate member support using a real knowledge base and escalating to humans when necessary. The same environment also monitors for churn signals in the base of serviced companies.
The strongest data point from this case is the scale:
- 35,000 companies under automated management with AI
- support with responses based on real knowledge
- escalation to humans when necessary
- behavior monitoring with churn risk alerts
There are no additional metrics in the fact pack regarding cost reduction, average service time, or conversion. Therefore, the case should be interpreted based on what is proven: large-scale operational use, not just testing.
2. Financial Market
In the financial market, Anthropic launched agents aimed at operators who perform specific intellectual tasks.
According to O Globo's coverage, these agents can:
- prepare presentations for client meetings
- review financial statements of companies under analysis
- escalate cases for compliance review
Here, the relevant data is not just the automation of repetitive tasks. It's the fact that agents are integrated into routines that require material reading, information organization, and the initiation of subsequent steps.
3. Mass Litigation
In the legal sector, Enter describes AI agents applied to mass litigation with a focus on operational scale and strategic decisions.
According to the company, these agents:
- read each document
- cross-reference external evidence
- detect fraud
- produce customized defenses and other legal documents
Enter's formulation is direct in separating two planes of use: a specialized AI for strategic decisions, with deep analysis and case-by-case adaptation, and another for operational scale, with standardization and end-to-end automated execution.
How AI Agents Work in Practice
The most objective definition from the material comes from Serasa Experian: autonomous AI agents are systems capable of perceiving the environment, deciding, and acting to achieve specific objectives. This helps explain why they appear in more complex processes than simple scripted service.
In practice, the functioning described by the sources combines several recurring elements:
- context input, such as documents, knowledge base, or behavioral signals
- defined objective, such as answering a question, reviewing a statement, or producing a document
- decision-making, including prioritization, routing, or escalation
- action execution, with response generation, analysis, or handover to a human step
The CDL Goiânia case clearly illustrates this flow: the agent accesses a real knowledge base, responds when there is sufficient basis, and escalates to humans when necessary. In Anthropic's case, the agent acts on work materials and escalates compliance situations. In the legal sector, Enter describes document reading, evidence cross-referencing, and customized document production.
The difference from conventional automations is less about “doing it alone” and more about acting based on variable context.
TI INSIDE Online reinforces this point by highlighting that AI agents need to understand the context to function effectively. Without this, the operation tends to revert to a fixed-flow pattern, with less adaptability.
The Impact on IT and Development Is Already Changing the Nature of Work
The adoption of agents not only alters business processes. It also changes the role of those who design, integrate, and supervise these solutions.
Receba os próximos artigos por e-mail
Conteúdo novo de Draivv direto na sua caixa de entrada. Sem spam.
According to Portal Information Management, so-called agentic engineering represents a redefinition of the role and professional value of software engineers and developers.
The portal's analysis is accompanied by an explicit warning about the technology's limitations. Silva, an AI specialist interviewed by the publication, states:
“It's about humans knowing how to leverage AI resources to deliver higher quality work, with greater agility and less prone to errors – not least because artificial intelligence is not infallible. It makes mistakes.” Silva, AI specialist, Portal Information Management
This statement is central because it shifts the discussion from generic enthusiasm to real-world operation. If AI makes mistakes, implementation needs to consider supervision, escalation criteria, and robust technical design.
The expansion of tools itself reinforces this scenario. Microsoft already offers resources for creating and using agents, and OpenAI launched GPT-5.5 in Codex, impacting programmers' routines and the development capacity of AI agents. The combined effect is clear: building this type of application is becoming more accessible, but this does not eliminate the need for architecture, validation, and human oversight.
What Makes Sense to Evaluate Before Applying Agents in Your Company
The examples in the fact pack do not authorize a universal list of best practices. But they do allow for the identification of concrete evaluation criteria based on already published cases.
1. Where a Usable Knowledge Base Exists
In the CDL Goiânia case, service stopped relying on fixed flows because it began operating on a real knowledge base. Without this input, the agent's autonomy tends to be limited.
2. Where There Are Tasks with a Clear Objective and Verifiable Output
The Anthropic and Enter cases show agents performing delimited tasks:
- prepare presentations
- review financial statements
- escalate compliance cases
- analyze documents
- cross-reference evidence
- produce customized documents
These are activities with a clearer beginning, middle, and end than open and diffuse demands.
3. Where the Process Already Allows for Human Escalation
In CDL Goiânia's service, the agent escalates to humans when necessary. This detail is important because the specialized source itself warns that AI is not infallible.
4. Where the Gain Is in Reducing Operational Bottlenecks
In all three cases, agents are integrated into operations with volume, repetition, or the need for structured analysis. The benefit described by the sources is not abstract: it is linked to the execution of real work.
Interpretation: What the Cases Show About Practical Application
Based on the documented examples, AI agents are advancing when a company needs to combine context, decision, and action within an operational process.
This interpretation stems from the presented cases:
- at CDL Goiânia, the agent responds, escalates, and monitors behavior
- at Anthropic, the agent organizes materials, reviews information, and escalates cases
- at Enter, the agent reads documents, cross-references evidence, and produces customized documents
It is also clear that adoption does not eliminate the human role. Silva's statement in Portal Information Management serves as a necessary counterpoint to technological advancement: AI can improve quality and agility, but it makes mistakes. Therefore, the design of the operation remains decisive.
What Comes Next
The next market milestones identified in the research material are in three areas:
- advancement of models and platforms that facilitate agent creation
- greater accessibility of corporate tools, such as the initiatives published by Microsoft
- evolution of AI-assisted development, with releases like GPT-5.5 in Codex
For businesses, the most concrete trigger is not to wait for a final market definition, but to observe where processes already exist with a knowledge base, delimited tasks, and the need for structured supervision. It is in this type of operation that AI agents are moving from promise to integrated intelligence.
If your company has already identified recurring bottlenecks in customer service, document analysis, or internal decision flows, the next step is to map where a custom agent can be safely implemented, with context and real productivity gains.
Related Content
Next Step with Draivv
Applying AI with results begins with choosing the right problem, data viability, and a clear business metric. Discover the AI for Business Diagnostic to transform scattered opportunities into a prioritized application roadmap.
