Will AI Agents Revolutionize How We Query and Use Data?

8 points by charliessnow ↗ HN
Snowflake just announced AI Data Agents in Cortex, a new way to automate and streamline data workflows with AI. Instead of just querying data or calling APIs, these agents can dynamically decide which tool to use—whether it's SQL, LLMs, or external APIs—depending on the task at hand.

This feels like a shift from traditional “query-driven” workflows to “goal-driven” automation, where agents choose the best approach rather than relying on static logic.

I’m particularly excited about building agents that select the right tool for the right job—instead of forcing everything into SQL, Python, or one ML model.

What do you think? Will AI-powered agents change how we interact with data? Or will traditional query-based approaches remain dominant?

https://www.snowflake.com/en/blog/ai-data-agents-snowflake-cortex/

7 comments

[ 2.1 ms ] story [ 33.3 ms ] thread
A few interesting technical aspects from diving into this:

The system uses Claude 3.5 Sonnet running inside Snowflake's secure perimeter for handling complex queries across both structured and unstructured data. This allows maintaining data governance while leveraging LLM capabilities.

Their search stack is using a hybrid approach combining vector and lexical search with semantic reranking. The benchmarks show a 12% improvement over OpenAI embedding models (measured on NDCG@10) using their Arctic-Embed-L-v2.0 model.

For structured data analysis, they're using semantic modeling rather than just pattern matching for Text-2-SQL generation. This helps handle complex multi-table environments and prevents common issues like JOIN hallucinations and double counting.

The quickstart [1] shows how to implement this in practice. Key components:

Agents that can plan multi-step queries Semantic SQL generation for structured data Hybrid search for unstructured content

There's a demo video [2] showing a practical insurance claims scenario that demonstrates:

How it handles complex document collections Real-time semantic understanding of data structures Multi-step query planning and execution Integration between structured DB queries and unstructured document search

Would love to hear from anyone who's worked on similar orchestration challenges across structured and unstructured data securely, and gain meaningful insights.What challenges have you encountered with mixing structured and unstructured data processing?

[1] https://quickstarts.snowflake.com/guide/getting_started_with... [2] https://www.youtube.com/watch?v=SDqyB-yWaAE

I see many people saying Anthropic has to release something new to compete, but I think sonnet 3.5 is doing just fine. That private companies can run it on AWS Bedrock and handle many of the GDPR concerns is something many people overlook. Also, if it is still producing the best code, why change it.
Agents are the way. Particularly with autonomous, unstructured workflows. It's all about tools. You give an agent a particular set of tools and a high level task, and that's it. No more complex logic to build, the LLM just figures it out.
I'm still trying to get my mind around the agent to agent task isolation and architecture. This seems to be the way forward, but will require retraining my brain to rethink how they form bonds vs. exist as atomic units.
Tool selection has been a tough nut to crack, especially when the “toolbox” gets really big.

How well does this work if we give it 10+ search services, SQL data models, python tools, etc. any benchmarks or first-hand experience?

>Tool selection has been a tough nut to crack, especially when the “toolbox” gets really big.

This really comes down to your tool design. It's more about crafting really high quality descriptions and examples versus worrying about any kind of prompting directly to the agent about usage. Make the tool highly self evident and the agent will naturaly use it.

The answer for every headline with a question is "no".