Workshop
The A³ Framework: Designing Agentic Systems You Can Trust, Govern, and Control
Room 32
11 November 2026, 12:30–13:30
As organizations move from AI assistants to increasingly autonomous AI agents, a new challenge emerges: how do we design systems that can act independently without losing transparency, control, and accountability? Many current discussions focus on models, tools, and frameworks. The architectural question is not which agent technology to use, but how to design agency itself: what decisions an AI system should make, when it should defer to humans, how multiple agents should collaborate, and how responsibility can be traced across the system.
This workshop introduces the A³ (Agency-Aware AI) Framework, a practical design approach for architects, developers, and technical leaders building the next generation of AI-powered systems. The framework provides a structured lens for engineering:
· Agency: defining goals, autonomy boundaries, and decision rights;
· Architecture: designing single-agent and multi-agent systems that coordinate effectively while remaining understandable;
· Accountability: embedding explainability, human oversight, uncertainty awareness, and governance directly into system design.
Rather than treating trustworthiness as an afterthought, A³ treats accountability as an architectural concern that shapes system behavior from the beginning. Whether you are building internal AI copilots, multi-agent workflows, decision-support applications, or enterprise AI platforms, the A³ Framework offers a practical method for balancing capability, control, and responsibility.
Laura Dioșan
Faculty of Mathematics and Computer Science
Laura Diana Cernău
Marsh | Babeș-Bolyai University





