Zürich AI | Agentic Autumn

Agentic Autumn will be hosted in partnership with mesoneer AG

Agenda
- Luca Campanella (CuratedAI) - Agents lessons learned
- Max Buckley (Google) - Agentic RAG

Abstract:
Luca Campanella (CuratedAI) - Agents lessons learned
We build narrow, citation-first systems for lawyers: CuratedAI automates EU privacy work (review, research anchored to EUR-Lex/EDPB with daily updates, and privacy-notice drafting from live site scans), while Praven Intelekt serves Bulgarian practitioners with retrieval-based answers, drafting, and document analysis—always linked to the exact article/paragraph and without using client uploads to train AI.
In the talk, I’ll share lessons: when to pivot vs. persist, sales and locality, in-person vs. remote, workflows vs. agents, why robust RAG is hard (many moving parts), and the value of a tight test set for search.

Max Buckley (Google) - Agentic RAG
Beyond Single Shot Retrieval: Agentic RAG Systems
Retrieval-Augmented Generation (RAG) has become foundational for building LLM applications that require access to external knowledge. However, traditional RAG pipelines — with their static retrieve-then-generate approach — struggle with complex queries that require multi-step reasoning, diverse information sources, or dynamic decision-making. This talk will explore the emerging paradigm of agentic RAG, where LLM agents orchestrate the retrieval process through iterative planning, tool use, and self-reflection. This enables query decomposition, sequential querying, and querying different tools and systems across diverse knowledge sources.