In the latest AI in action sessions, Urudata Software’s internal knowledge-sharing series, members of the technical team shared new ways to leverage artificial intelligence in their day-to-day work. This time, the presentations focused on two areas of software development: code review and project management.
Modular Code Review: make your agent an expert in your product
The session, led by Nicolás Dodino, Intermediate Developer, explored code review with AI agents and how to improve their results through configurations tailored to each product.
When an agent receives a pull request without sufficient context, it may generate overly generic feedback or apply rules that do not match the application’s specific requirements. To address this issue, Nicolás presented an approach based on modular skills that organize review rules according to different product areas, such as frontend, backend, databases, and testing.
Instead of loading all the rules into a single file, the idea is to identify which modules have been affected by the changes and use only the relevant criteria for each review. This allows the agent to work with more specific instructions, avoid unnecessary checks, and focus on the risks associated with each modification.
Another important aspect is classifying findings according to their severity. Distinguishing a critical error from a minor observation makes it possible to prioritize issues that could actually compromise a delivery. The proposal also included adding a category for the agent’s questions, allowing ambiguous cases to be evaluated by the team before a final verdict is issued.
The demonstration also covered the detection of potential security issues, vulnerable dependencies, and changes requiring special attention.
AI Boards: Azure Boards in natural language
In the most recent session, Andrés López, Semi Senior Developer, presented an integration between Azure Boards and AI tools that makes it possible to interact with project tasks using natural language.
Through a connection with Azure DevOps, users can check assigned tasks, view their status and priority, request summaries of completed work, and manage pending items without constantly switching between environments. The tool also allows users to request actions on tickets, such as updating their status or identifying completed tasks that have not yet been closed.
One of the most interesting use cases comes up when a task contains all the information needed to develop a feature. In this case, an agent can be asked to pick up the ticket, carry out the development, and create a pull request, recording the link so the team can focus on reviewing the code and validating the outcome.
The integration also makes it possible to organize parallel work through subagents, assigning one task to each and generating a separate PR for every task. AI can therefore contribute to managing pending work and handling certain stages of development, while the team remains responsible for reviewing the changes and ensuring they meet the client’s requirements.
We keep putting AI into action
AI in action continues to provide a space for sharing knowledge, experimenting with new technologies, and exploring ways to apply them to Urudata Software projects. We keep testing new tools, exchanging experiences, and building solutions together.