We use cookies to optimize your experience, analyze traffic, and personalize ads. Please choose whether you accept our use of non-essential cookies. Read More
The race to adopt generative Artificial Intelligence (AI) has placed technology leaders and executives before an urgent dilemma: how to connect traditional enterprise applications, often built decades ago, to advanced AI models without compromising operational security or generating unpredictable costs?
Plugging AI directly into a legacy ERP database without prior processing does not produce innovation. In practice, this approach creates biased decisions, exposes critical information, and compromises return on investment (ROI).
This was the central theme of an in-depth discussion on the Builders by Skyone, led by Robson Del Fiol, Director of Education at Skyone, and Luciano Condé, Senior Director of Architecture at Oracle, professor and researcher with over 35 years of experience in the technology market.
Drawing on the practical experience of someone who has witnessed the evolution from the opening of the IT market reserve in Brazil to the consolidation of cloud computing and large language models (LLMs), this dialogue demystifies the hype surrounding AI and presents a technical roadmap for modernizing operations securely.
There is a common misconception in the technology market that legacy systems are synonymous with being outdated. For established companies, these applications represent decades of validated business rules, mature processes, and, above all, a rich history of data.
“Having legacy systems isn’t a bad thing. You have a history there: you have data, culture, and processes. The fundamental challenge is how you accelerate that legacy and place it within a fast, scalable, and, above all, robust reality.”
says Luciano Condé, Senior Director of Architecture at Oracle
Modernizing these structures does not necessarily require a complete rewriting of complex codes—a move that is usually expensive and risky[cite: 1, 3]. Architectural evolution involves three complementary fronts:
With platforms like Skyone Autosky, for example, companies can migrate their ERPs to the cloud with Zero Trust architecture, minute-by-minute autoscaling, and complete network isolation, creating the necessary secure foundation even before starting artificial intelligence projects.
To make assertive investment decisions, managers need to understand that Artificial Intelligence did not emerge with recent conversational interfaces. As Condé reminds us, the conceptual foundations of the field date back to the Dartmouth College seminar in 1956, with the goal of mimicking human cognitive capabilities.
The practical evolution of technology is divided into two major waves that complement each other in companies today:
"When you type something into a generative AI, it's not imagining or thinking like a human; it's calculating the mathematical probability of the next word from a previously trained mass of data."
explains Luciano Condé
Understanding this mechanism shatters the illusion that AI solves problems on its own. Without organized and contextualized corporate data, any generative model will deliver generic or incorrect answers.
One of the most frequent mistakes in corporate AI adoption is connecting models directly to production databases without intermediate processing and control layers. This practice compromises the project on three critical fronts:
Granting unrestricted read and write access to an autonomous agent over a production database opens serious vulnerabilities. Without clear permission and isolation policies, a misinterpretation can result in the accidental deletion of records or the leakage of sensitive company information.
Corporate data carries historical distortions. If a manager feeds an AI only the last 12 months of sales to generate a forecast,the model will ignore longer economic cycles, atypical seasonality, or critical external events that distorted that specific period.
The quality of AI-generated analysis is directly proportional to the quality of data hygiene in the input. Duplicate databases, non-standardized tables, and incomplete records produce misleading information and reports that lead management to make mistakes.
To transform legacy data into a real competitive advantage, integration needs to be mediated by a solid governance architecture. During the conversation, Robson Del Fiol and Luciano Condé highlighted three essential technical pillars for this operation:
| Architectural Pillar | Strategic Function | Business Impact |
| iPaaS Platform | It orchestrates communication between ERPs, databases, APIs, and AI models. | It centralizes access rules, defines who can access each piece of data, and avoids weak, manually implemented integrations. |
| RAG (Retrieval-Augmented Generation) | It connects the language model to a private, sanitized company knowledge base. | It ensures that the AI responds based on real business facts and documents, reducing misinterpretations. |
| Human-in-the-Loop | It maintains mandatory human validation at critical points in the automated workflow. | It prevents AI from performing irreversible actions (such as deleting data or issuing invoices) without approval. |
Luciano Condé points out
“An iPaaS platform allows you to establish governance: when, who, and how each system can access the data. And when it comes to taking critical actions, you need human-in-the-loop, the active participation of a person validating the process so that the AI doesn't go haywire in sensitive operations.”
To demonstrate how this architecture works beyond theory, Robson and Condé held a live vibe coding, an approach in which professionals describe objectives in natural language so that AI agents can structure functional applications on real databases.
Using Skyone Studio Creator connected to an ERP test environment (running on Oracle infrastructure), the experiment demonstrated the rapid creation of solutions with governance:
Even immersed daily in complex cloud architectures, algorithms, and doctoral research in B2B strategy, Luciano Condé concluded the meeting by sharing an analog habit that has sustained his mental clarity for over five years: writing three handwritten pages every morning before touching his cell phone or computer.
The practice functions like a download . By transferring concerns, ideas, and perceptions to paper immediately upon waking, the executive identifies recurring thought patterns and makes conscious decisions before the operational routine takes over the day. In a market accelerated by artificial intelligence, the capacity for structured reflection remains the main human differentiator.
This article summarizes the key technical insights from the episode featuring Luciano Condé (Oracle) and Robson Del Fiol (Skyone), but the full conversation offers even more behind-the-scenes insights into the history of computing in Brazil, prompt engineering (such as the CRISP framework), and the real challenges of digital transformation.
Press play and listen to the full episode of the Builders by Skyone podcast on Spotify now to catch all the details of this conversation and discover how to prepare your company to scale safely in the age of Artificial Intelligence.
Transform Your Business with Skyone. Request a demo or schedule a call with our experts to discover how Skyone can accelerate your digital strategy.
Have a question? Talk to a specialist and get all your questions about the platform answered.