SQL Functions Are Quietly Becoming the New AI Interface

Databricks exposing document extraction and classification as plain SQL functions — ai_parse_document, ai_extract, ai_classify — is a small technical detail with a bigger pattern behind it: AI capability is steadily moving into the exact tools teams already use, rather than requiring a new specialist tool per capability.

Why This Pattern Matters More Than Any Single Feature

Every time an AI capability becomes callable from an interface people already know — SQL for analysts, spreadsheet functions for finance teams, chat for knowledge workers — the practical barrier to adoption drops sharply. The capability existed before; what changed is who can actually use it without new training or a new tool procurement cycle.

Where to Look for This Pattern in Your Own Stack

• Check whether your existing data platform (Databricks, Snowflake, BigQuery, or similar) has added AI-native SQL functions you haven't audited yet — these often ship quietly, without a major product announcement
• Look for AI capability appearing inside tools your non-technical teams already use daily — that's usually where adoption actually happens, more than in a dedicated AI tool few people open
• Evaluate new AI tooling purchases against the question 'could this capability instead live inside a tool we already have?' — sometimes yes, which changes the buy-vs-use-what-you-have calculus

The Skill Implication

As AI capability moves into SQL, spreadsheets, and other familiar interfaces, the actual skill gap shifts from 'who can build an ML pipeline' to 'who can write a good query and interpret the result' — a much larger pool of people inside most organisations, which is exactly the democratisation effect worth planning your training investment around.