Software Architecture in the Age of Agentic AI
Authors: Karthik Vaidhyanathan (SERC, IIIT Hyderabad, India) and Henry Muccini (University of L’Aquila, Italy)
The increasing use of generative AI and agent-based systems in software development has led to new challenges and opportunities in software architecture. Two broad areas of research are emerging: i) software architecture (SA) for building AI-enabled systems, including those with LLMs and autonomous agents; and ii) the use of AI techniques, especially generative models, to support architectural activities such as decision-making, documentation, etc. While both directions are gaining interest, contributions are scattered across different research communities including software architecture, AI/ML, and self-adaptive systems. This tutorial aims to bring together ideas and experiences from both research and practice in a structured way. The tutorial is divided into three parts: i) SA for ML-enabled systems (SAML), ii) LLMs for supporting software architectural processes and practices (LLM4SA), and iii) Architecting multi-agent systems with generative AI (Agentic AI). Each part will include key terminology, research themes, example systems/demos/case studies, and hands-on/discussion activities. The goal is to help participants understand the current state of research, share practical tools and frameworks, and identify future research directions in this fast-evolving area. .