درباره این اپیزود
This is a guest conversation episode of Ship It Weekly, separate from the weekly news recaps.
In this Ship It Conversations episode, I talk with Tristan Handy about what changes when AI agents, not people, become the main consumers of your company's data, why the hard part isn't text-to-SQL, and why he thinks the most interesting work is operating on the system instead of inside it.
Tristan is a co-founder of dbt Labs and now president of Fivetran + dbt. If you've worked anywhere near a data team, you've probably run into dbt. His team likes to call it Terraform for data.
We start with his line that dashboards are a hack for the human visual cortex. As agents take over the pattern-finding, the asset moves down the stack: a well-modeled gold layer, a semantic layer that captures how your organization calculates its metrics, and a context store that helps the agent make sense of all of it.
On reliability, Tristan says agents already write SQL that compiles and runs almost every time. The hard part is whether they understood the question and the business behind it. In the benchmarks he's seen, an agent turned loose on raw enterprise data gets about 30% of its answers semantically right. Give it context and a semantic layer, and it's north of 90%.
We also get into where autonomy should stop once agents can act on data, why analytics engineers (and SREs) should be operating on the system rather than inside it, how dbt earned practitioners' trust, what he got wrong selling to enterprises, and where dbt draws its open source line after the Fivetran merger.
His first step for a platform or data team is to eat your spinach: a well-governed data lake, well-modeled data and a semantic layer, then a context layer and evals. And his team paused dbt 2 work for two months to build an agent loop that took conformance from about 45% to 93%.
Highlights
• Why dashboards stop being the product when agents read the data
• Gold layer, semantic layer and context store
• Text-to-SQL vs semantic correctness: about 30% vs north of 90%
• Where useful autonomy ends and regret begins
• Operating on the system instead of inside it
• How dbt earned trust, and what it got wrong selling to enterprises
• State and UI as the open source line
• Eat your spinach, then a context layer and evals
• A software factory that got dbt 2 conformance from 45% to 93%
Links
Tristan on LinkedIn · dbt · Fivetran · The Analytics Engineering Roundup · The Analytics Engineering Podcast · Spider 2.0 benchmark · More episodes · On Call Brief · LMGT Awards
رونوشت 🔗
00:00:00.020 --> 00:00:04.179
Dashboards are a hack for the human visual cortex.
00:00:04.500 --> 00:00:08.480
That's how my guest today puts it. They pipe
00:00:08.480 --> 00:00:12.779
a lot of information into our brains fast so
00:00:12.779 --> 00:00:16.199
we can spot the patterns. But more and more,
00:00:16.359 --> 00:00:19.760
the thing reading your company's data isn't a
00:00:19.760 --> 00:00:24.199
person. It's an AI agent. And agents are already
00:00:24.199 --> 00:00:27.899
gr…