Agentic AI is moving from experimentation into production-grade testing, even in highly regulated banking environments. This session shows how agentic testing is already being applied in practice, building on LGT’s award-winning work with Infometis and the ongoing evolution of the Avalon Agent.

The talk demonstrates how agents use context grounding and Retrieval-Augmented Generation (RAG) over bank-internal test artefacts and systems, while Model Context Protocol (MCP) servers expose deterministic, auditable tools for agents to interact with enterprise systems in a controlled way. Rather than another AI vision talk, this session focuses on what already works, where the limits are, and how to operationalize agentic testing in banking.