“AI is here to stay. We need to understand how to live with it and how to leverage it.” — João Silva, Head of Engineering at Consulteer Portugal
AI is already changing how software teams work. At Consulteer Portugal, this became more than a topic of discussion. It became a summer internship project.
Mentored by João Silva, Head of Engineering at Consulteer Portugal, interns Ana Caroline Soares Silva and João Pedro Nunes Ferreira explored how AI could support a challenge many software teams know well: understanding and documenting existing codebases.
Legacy systems often contain valuable business logic. But documentation is not always complete, current or easy to access. Before teams can improve a system, they first need to understand it.
That was the starting point: could AI help create a first layer of orientation?
The Use Case.
“The interns built a workflow using AI models to help document legacy code.” — João Silva
The project worked with code from GitHub repositories — the kind of codebase teams often need to understand before they can improve it. The question was simple: could AI create a first layer of documentation and help teams get oriented faster?
Ana Caroline and João Pedro tested two different routes: one using low- and no-code tools, the other with a more code-based setup. Comparing both approaches made the project useful. It showed where AI can speed things up, where human review is still needed, and which setup might work best in different situations.
Where it clicked.
“When I tested the agent on a large university project, the result was better than I expected.” — João Pedro Nunes Ferreira
For João Pedro, the project clicked when he tested the AI workflow on one of his own university projects.
The output was not something to use blindly. It still needed review, context and human judgement. But it gave a useful first structure — and showed how AI can help teams get started with unfamiliar code.
That was the shift: AI became less of an abstract topic and more of a tool he could test, question and improve.
The Reality Check.
“The toughest part was getting my project to run in a self-hosted environment.” — Ana Caroline Soares Silva
Applying AI in a real setup is rarely as simple as the demo suggests.
For Ana Caroline, one of the biggest challenges was getting her project to run in a self-hosted environment. That meant dealing not only with the tool itself, but also with infrastructure, access, dependencies and security requirements.
It made the project more realistic and showed what happens when a promising idea meets real technical constraints.
Why it matters.
“AI can save time on repetitive tasks, so teams can focus more on creative work.” — Ana Caroline Soares Silva
The project was not about proving that AI can solve documentation on its own.
Its value was more practical: AI can help with the first layer of orientation. It can reduce repetitive groundwork, create structure and help teams move faster when they face an unfamiliar codebase.
The output still needs review. Context still matters. People still make the decisions.
But if AI supports the first step, teams can spend more time on the work that needs human judgement.
Learning by doing
“The people are really friendly. I felt comfortable trying, failing and learning every day.” — Ana Caroline Soares Silva
For Ana Caroline and João Pedro, the internship was not only about AI. It was about learning how real project work happens.
They tested ideas, compared approaches and worked through technical limits. Some things worked quickly. Others needed troubleshooting. That was the point.
The project gave them room to ask questions, try things out and learn from what did not work immediately. And that made the experience feel closer to real engineering work than a polished demo ever could.
This keeps the focus on the interns and avoids repeating the “don’t be afraid of AI” message too often.
What’s next?
“Go ahead, do what you have to do, and you will enjoy the process.” — João Pedro Nunes Ferreira
The internship did not end with all the answers. It ended with a better understanding of the questions.
What can AI take off a team’s plate? Where does human judgement become essential? And how do different approaches change what is possible?
For Ana Caroline and João Pedro, those questions became concrete through practice. They tested, adjusted, failed, learned and saw how AI can fit into real software work.
For Consulteer Portugal, the project opened a useful path: keep exploring AI where it meets actual project challenges.
That is what stayed after the summer: not a finished solution, but a clearer view of what could come next.
Want to become a Consulteer?
Explore our open positions. Your next opportunity might already be waiting.
