Why I'm Actually Excited About AI-Native Data Workspaces (And Why I Wasn't Before)

Last month, I spent three hours debugging why an analyst on my team kept getting stale data. Turns out they were running queries in Snowflake, copy-pasting results into a Slack message, and I was reading those Slack messages to build a report. We had a perfectly good data warehouse. We just didn't have a shared place to think about it together.

That's when I realized something: we've solved the data storage problem brilliantly. Snowflake, BigQuery, and friends are genuinely excellent at scale. But we've barely solved the collaboration problem around data. Most data teams are still working like it's 2015—SQL in one tool, results in another, AI suggestions in a third, and nobody's really looking at the same thing anymore.

So when I saw that Dory added Snowflake support as a "first-class data source," my cynical take softened a bit. This isn't just another SQL editor bolting on database connectivity. This is someone thinking about the actual workflow we're broken at.

What Dory Actually Built Here

Dory's approach is refreshingly straightforward. You connect your Snowflake account, and then you get a SQL console, schema browser, table previews, and a workspace where queries and their results live together. No copying. No switching between five applications.

The authentication design caught my attention immediately. They support both password and key-pair auth, but here's what matters: they embed the private key content directly instead of relying on local file paths. That's a small decision with big implications. It means the same connection works on your desktop and a self-hosted web deployment without configuration gymnastics.

I've fought with tools that ignore deployment differences before. This feels like it was built by someone who's actually shipped to production.

Why This Matters More Than It Sounds

Data teams aren't monolithic anymore. You've got analysts who want to explore, engineers who want reproducibility, operators who are losing sleep over secret management, and increasingly, AI agents that need structured access without turning database work into a chat transcript you can't audit.

Dory's workspace model tries to hold all of that together. A human and an AI can work against the same Snowflake context. The human can see exactly what SQL was executed. They can modify it, rerun it, audit it. That's meaningful because it keeps humans in the loop—not as an afterthought, but as part of the process.

My Take: Smart Foundations, Obvious Gaps

I like the intentionality here. The first release doesn't try to boil the ocean. No SSO, OAuth, MFA, or token caching yet. They prioritized portability and predictability over feature completeness. That's the right call for a v1.

But I also notice what's missing, and it matters for real organizations. MFA support feels non-negotiable at most companies I work with. Workload identity federation (for cloud-native deployments) would be huge. And token caching could make a real difference for performance-sensitive workflows.

The bigger question I have is adoption friction. SQL editors and data exploration tools are crowded. Dory needs to convince teams that AI-native collaboration is worth migrating their workflow. That's not a technical problem—it's a culture and momentum problem.

What This Means in Practice

If you're managing a Snowflake instance with a team that constantly fights context switching and audit trails, this is worth a genuine pilot. The key-pair auth implementation alone tells me the builders understand production constraints.

Here's a basic connection validation you'd run:

select 
  current_database() as database_name,
  current_schema() as schema_name,
  current_role() as role_name,
  current_warehouse() as warehouse_name,
  current_timestamp() as connection_test_time;

If that query returns cleanly, your connection is solid and you're ready to explore your schema hierarchy through Dory's browser.

The real test isn't whether the tool connects to Snowflake—every tool does that now. It's whether the workspace model actually changes how your team collaborates. Does having queries, results, and AI suggestions in one place actually reduce friction? Or does it just add another tool to your rotation?

What I'm Watching For

I want to see how this evolves. Will they add OAuth and MFA without losing the simplicity that makes the current release work? Will the AI agent workflows actually be auditable, or will they turn into black boxes? Will teams actually migrate to this, or is it just another thing everyone tries for a week?

I'm also curious whether Dory will expand to other data platforms. Postgres, BigQuery, DuckDB? The architecture seems flexible enough. If they nail the human-AI-database collaboration model, the database choice becomes almost secondary.

For now, I'm genuinely interested. Not because it solves an impossible problem, but because it's solving a real one that most of us have just accepted as broken.

Source: This post was inspired by "Dory Now Supports Snowflake" by Dev.to. Read the original article