Convesational Analytics: Building AI-Powered Analytics People Can Actually Trust

For decades, data teams have chased the dream of true self-service analytics: where stakeholders can get their own answers from curated datasets. With today's AI systems, that dream is finally becoming possible but turning a prototype into a trusted production system is harder than it looks.

This talk will distill what leading data teams are doing as they build conversational analytics in the real world. We'll look at the two features doing the real work under the hood: the semantic layer and vector indexes, and how together they turn a database into something you can actually have a conversation with. We'll cover practical, available techniques including how teams can use dbt to manage the semantic layer and vector index capabilities built into platforms like Snowflake and BigQuery. As well as the foundational work that doesn't go away: clean models, clear documentation, strong data quality, observability, and semantic consistency.

We'll also dig into why trust depends on more than model accuracy alone, evals, auditing, public feedback loops, and the changing role of data teams as conversational interfaces become a core part of the analytics stack.

If you want to understand where conversational analytics is today, what separates demos from durable systems, and how to prepare your data team for this shift this session will give you a practical foundation.