DataFramed

#376 Rethinking the Data Stack in the age of AI with Tristan Handy, President of Fivetran + dbt Labs

42 min

DataFramed Episode #376 features Tristan Handy, President of Fivetran + dbt Labs, discussing the evolution of the data stack in the age of AI. The episode covers the rationale behind the Fivetran and dbt merger, the rise of open data infrastructure, and how high-quality data management and semantic layers are key to enabling conversational analytics and agentic AI use cases. Handy also shares insights on how data teams are restructuring and the importance of becoming a full-stack data practitioner.

Summarized by Podsumo

🎧 Listen 🎙️ Ask about this episode

✨ Key Takeaways

1

The merger of dbt and Fivetran aims to provide an integrated, open data stack built around Iceberg and open table formats, offering customers choice and reducing vendor lock-in.
2

High-quality data management, clean gold-layer tables, and machine-readable documentation are essential for enabling accurate "conversational analytics" with AI agents.
3

A governed semantic layer (e.g., Apache Aussie) is critical for AI agents, providing canonical definitions of metrics to improve accuracy, reduce latency, and lower token costs.
4

Data teams are bifurcating into platform maintenance and democratization/enablement, with a growing need for full-stack data practitioners who can own projects end-to-end.
5

Handy emphasizes using AI agents as research assistants, providing specific tips like breaking down research steps and requiring proof at every stage to avoid hallucinations.

💬 Notable Quotes

"It's like having a team of graduate assistants that you're working with all the time, and you get to be the primary investigator, as opposed to you have to go and read a bajillion documents and synthesize them all in your head."
"The customers table that one team uses and then the customers table that another team uses, and there's like the one yes no flag that indicates this thing and the other yes no flag that's kind of almost the same thing, but it's just a little bit different."
"I don't think it makes sense generally for business users throughout the organization to be just creating metrics willy-nilly because that will actually lead to a bunch of sprawl and confusion and the quality of your answers are gonna go down."