ITSS Innovation Circle #1: Diving into the world of AI Agents

Author: Radu Bobe - Digital Identity Consultant

Last week, we hosted the first ITSS Innovation Circle: an interactive workshop designed to spark innovation across our teams. As in 2025 we continue our AI journey, the topic of the series is centered on AI Agents. For over 2 hours, our colleagues explored how Agentic AI is evolving from a "cool tool" to a real digital teammate, driving productivity and efficiency across various domains. Because the discussions were so engaging, in this blog post we would like to share a brief recap of the event with you.  

Workshop insights 

The session kicked off with a deep dive into the central concept of the workshop: AI Agents.  

Our speakers highlighted that nowadays an AI system does not just respond to prompts, but acts with autonomy, adaptability, and purpose. Certain examples of AI Agents were offered, presenting both the advantages and the current challenges, such as ethical constraints and limited real-world autonomy.  

Next, our colleagues presented the AI agents developed by each IT Smart Systems department over last months, sparking a true brainstorming session. The first one was the AI-Powered Troubleshooting Agent for Digital Identity, an AI Agent for troubleshooting identity platforms like Ping or Keycloak. As manual troubleshooting is time-consuming, error-prone, and heavily dependent on experts, we consider that our solution using AWS Bedrock acts like a virtual engineer, being capable of scanning logs, interpreting configs, searching documentation and, of course, proposing the fixes.  

The workshop continued with the presentation of the Payment Volume Monitoring Agent, a comprehensive tool for analyzing and monitoring banking API logs, designed specifically for the PSD2 landscape. The Agentic AI system automatically analyzes logs from banks and generates detailed reports and visualizations on API performance. The system drastically reduces issue analysis time from hours to minutes by  delivering actionable insights into banking integration performance. 

Last but not least, we discovered the Project Guardian: Code Review Agent for GitLab, an AI assistant that automatically reviews code. The solution identifies bugs, security flaws, and inefficiencies, providing inline comments with suggested fixes. In this way, our developers benefit from faster reviews, fewer bugs, and less manual overhead, leading to faster delivery and higher-quality code. 

It is important to mention that each Innovation Circle workshop will also include a strong hands-on component. In this first session, our colleagues took their first steps in setting up a working environment tailored for AI agents and began exploring how to interact with a large language model (LLM). 

As our engagement with Agentic AI is ongoing on a daily basis, please stay tuned for more technical insights into our AI agents, as well as updates from future Innovation Circle sessions!