Author: Kevin

AI is extremely fast.AI is very confident.AI will also look you straight in the eye and say, “Yes, this query is correct,” while quietly summing revenue twice and joining on vibes. This is where Human-in-the-Loop (HITL) comes in — the idea that sometimes, just sometimes, an actual human should be involved before AI does something……

AI is amazing.AI is powerful.AI will also happily spend your entire Snowflake budget in one afternoon if you don’t put it on a leash. Welcome to AI cost control – the phase of the journey where leadership stops asking “What can AI do?” and starts asking “Why did our credit usage look like a Motley…

AI in Snowflake is powerful.AI in Snowflake without guardrails is how you end up explaining to security why an LLM just discovered a table that it absolutely should not have. Think less “Welcome to the Future” and more “Everybody Wants to Rule the World” — until someone forgets who’s actually in charge. This post is…

Part of the Snowflake + AI Practical Series That Keeps Getting Spicier Let’s be honest: If SQL queries were people, half of them would show up to work late, wearing mismatched shoes, smelling faintly of cold brew and guilt. And yet… we trust these things with our financials, our KPIs, our CEO’s dashboard that must…

If you’ve ever asked an LLM to “just write the SQL,” you already know what horror feels like. The query looks fine at first glance… until you discover a join that makes no sense, logic that defies basic business rules, and a confidence level that suggests the AI should lie down for a bit. Welcome…

Part 2 of the Snowflake AI Query Reliability Series If Part 1 was all about protecting yourself from questionable AI-written SQL, Part 2 is where you stop reacting and start training your model to behave. AI will happily generate SQL with the confidence of a ’90s action hero and the accuracy of someone reading your…

AI-generated SQL is everywhere now — especially inside Snowflake.But let’s be honest: sometimes AI writes queries with the same confidence (and accuracy) as Vanilla Ice claiming he didn’t sample Queen. If you want accurate, reliable SQL from LLMs, you need structure. Not vibes. Not hope. Actual Patterns. The kind of repeatable patterns that keep Snowflake…

If the last post was about using AI to argue with your SQL, this one is about using AI to negotiate with it — the way JR Ewing negotiated: confidently, with a smirk, and fully prepared to expose what someone doesn’t want you to see. Welcome to the next chapter in this Snowflake + AI…

If Part 1 was about building trust, Part 2 is about giving your AI coworkers permission to bicker — productively. Because if “trust, but verify” was the Cold War version of collaboration, “let the machines argue” is the modern data equivalent. You don’t want a single all-knowing LLM nodding along; you want a tiny debate…

Part 1: Why AI-Validated Queries Are the New Data Governance Somewhere between a DBA whispering “it worked on my dev box” and an analyst insisting “the dashboard’s fine, I swear,” lies the truth — and usually, a join that went awkwardly wrong. In the age of AI, we’re generating queries faster than ever… and making…