AI-powered test automation: AI Fluency culture in practice
In software engineering, the gap between theory learned in universities and practice when facing complex code can be daunting. When the urgency for innovation knocks on the door, the fear of “I don’t know where to start” often paralyzes even the most dedicated professionals.
Today, we’re here to tell a story that proves that technical humility, business vision, and a culture of AI fluency is a combination capable of transforming vulnerabilities into career leaps.
Suelen Lemos, a software quality professional at Squadra, transitioned from the support team to the testing team and took on the challenge of structuring test automation for a critical system from scratch. In addition to accelerating her professional development, her project served as inspiration for the creation of a turing bot to support numerous other QA initiatives at Squadra.
The starting point
Suelen began her journey in the field as a support professional and, due to her quick grasp of business rules, moved to the manual functional testing team. However, as the application grew, manual execution of scenarios became a bottleneck:
“I was faced with a growing volume of manual scenarios and less and less time to execute them with the necessary quality. It was in this context that I understood that automation would be the way to gain efficiency, optimize my time, and increase test reliability.”
Despite having completed postgraduate studies and several theoretical courses in the area, in practice she encountered a highly complex, critical system with no formal documentation of business rules: Geosite Telecom, a platform based on geolocation and used by over 500 internet providers to manage all documentation and maintenance of fiber optic networks (FTTH).
Suelen needed to create the platform’s test automation from scratch, defining architecture, standards, and scope, all without practical experience:
“Although I had theoretical knowledge in automation and practical experience with Cypress on static web pages, I had never structured an automation project from scratch, nor did I have experience with Playwright. Faced with the system’s complexity, I realized I would need support to take the first step. The best decision I made was to ask for help, without letting pride or the fear of ‘I don’t know’ paralyze me.”
AI as a code generator and individual mentor
With the support of an expert, Stenio de Abreu, Head of Chapter at Squadra, Suelen connected Claude Code to the Geosite Telecom repository. The major turning point in her technical evolution occurred in how she began to interact with AI. Instead of just using it to write code, Suelen adopted an investigative approach. For every suggestion the AI made, such as the choice of the Playwright framework, the rule was always to ask why:
“Claude Code went far beyond generating code: it understood how the project’s tools worked and helped build complete strategies. This habit of always asking ‘why’ behind each decision is what transformed the tool into a true learning tutor.”
This dynamic turned the code-writing process into continuous, one-on-one mentoring. The project was born, and Suelen deeply understood the function of each layer, each file, and each of the best practices applied.
The irreplaceable human protagonism: when business knowledge guides AI
Suelen’s evolution was not driven by blind submission to AI, but by affirming her own critical role as an expert in the system’s business rules. On several occasions, the AI generated syntactically correct, but functionally inconsistent code and assertions with the real world. It was Suelen’s keen eye that identified and corrected these deviations:
“There were situations where the AI would complete the registration of a piece of equipment in a completely out-of-context location, like in the middle of a forest. In these cases, my role was to identify the inconsistency and adjust the automation so that the registration respected the real business rules. This reinforces that AI accelerates the process, but human judgment remains irreplaceable.”
Suelen’s mastery of the processes allowed her to guide the AI in scenarios where there was no formal documentation. Throughout the project, she not only automated the system’s testing structure but also helped document the functional rules that the application had never had recorded.
The methodology developed throughout this project revealed a replicable pattern for any professional or squad facing the same challenge, consolidating the creation of an expert turing bot:
“It’s gratifying to see that my experience contributed to the development of a turing bot, a tool that can be applied by anyone within Squadra.”
Beyond the learning curve, exponential efficiency gains
The combination of the courage to learn and AI support drastically reduced the time needed to master automation. The project’s final results illustrate this acceleration by comparing the following scenarios:
Beginner QAs without senior mentorship: 2 to 4 weeks, with the risk of not adopting best practices due to lack of experience;
Beginner QAs with senior mentorship: 1 to 2 weeks;
Beginner QAs with AI and senior mentorship: 4 days.
With focused work and accelerated learning, Suelen delivered a robust, scalable testing architecture ready to run continuous integration pipelines.
This story is a watershed moment in the culture of AI fluency driven daily at Squadra. It demonstrates that seniority and technical mastery in the AI era are not born from knowing everything beforehand, but from the willingness to learn with curiosity and humility. For everyone who feels hesitant to take the next step, Suelen leaves a message:
“Humility to say ‘I don’t know yet, but I will learn’ is the first step. Eliminate the belief that you need to master everything before starting. It was this combination of asking for help, maintaining curiosity, and trusting the process that brought me here, and I’m sure it can serve as inspiration for others.”
