While models and tools for building AI agents continue to evolve quickly, the long-term success of agentic solutions depends far more on sound design, architecture, and user experience than on any single framework or SDK.

In this session, you will learn how to design AI agents and agentic AI systems that are reliable, understandable, and usable in real-world enterprise environments. We will focus on the architectural and UX decisions that are most critical when building agents that reason, plan, and act on behalf of users.

The session will cover core agent design concepts such as defining agent responsibilities and boundaries, choosing between single-agent and multi-agent architectures, managing context and memory, and designing action and tool interfaces that are safe and predictable. You will explore patterns for orchestration, human-in-the-loop workflows, and failure handling, along with common anti-patterns that lead to brittle or untrustworthy agents.

Beyond architecture, the talk emphasizes agent user experience, how users interact with agents, how agents explain their reasoning and actions, and how to design conversational and non-conversational interfaces that build trust and clarity. Real-world enterprise scenarios will be used throughout to illustrate how these design choices affect maintainability, governance, and adoption.

Attendees will leave with a practical mental model and a set of design patterns they can apply immediately when architecting AI agents inside modern applications.

You will learn:

• Understand the core design principles behind agentic AI systems
• Architectural patterns for single-agent and multi-agent solutions
• User experience principles that improve transparency, trust, and usability in AI agent interactions