Data Modeling is Dead (Again), 2026 Edition — Part 2

Part 2 of Reis’s rebuttal to the “data modeling is dead because AI” argument. Rather than point-by-point refutation, he frames it through Joel Spolsky’s Law of Leaky Abstractions: increasing abstractions doesn’t remove complexity, it forces you to understand it deeper.

Core argument: an AI agent can generate syntactically perfect SQL schemas but lacks context about arcane business operations (e.g., how a logistics partner splits shipments, or that “customer” is legally distinct in the EU vs. US). The model looks convincing but is wrong. Hiding complexity with AI just obscures failure points.

Reis concedes exceptions for prototypes and weekend projects, but for production systems handling operations, revenue, and lives, skipping data modeling with agents is reckless.

RDCO relevance

Strong alignment with our consulting positioning — we help clients who’ve tried the “just let AI generate it” approach and hit the wall. The Spolsky framing is a useful reference for client conversations about why dbt modeling discipline still matters in an AI-accelerated world.