SQL is Dead. Long Live SQL.
AI just killed SQL. Or did it?
After fifty years of faithful service, is SQL quietly writing its obituary, one natural language query at a time? Industry experts like Bruno Aziza from IBM and Dave Mariani from AtScale suggest users no longer need to write SQL. Instead, natural language interfaces, powered by semantic layers, translate business questions into data retrieval logic, making SQL coding obsolete.
To grasp the impact of this claim, consider a marketing analyst who needs to understand recent customer churn trends in her company.
Instead of asking a data engineer to hand-craft complicated SQL queries to join multiple tables, she now crafts a rich prompt for AI. She can ask, in plain English: "Show me the top 5 reasons enterprise customers churned last quarter, correlated with support ticket volume. Check trends for each reason over the last three months, and draft a report summarizing your findings."
The AI reasoning model interprets her question. It asks the semantic layer to map business terms like "Customers" or "Ticket Volume" to underlying physical database structures. Data is retrieved and integrated into the AI workflow. The user gets an answer. Effectively, this new workflow eliminates the need for humans to write SQL. The new paradigm seems clear: "Ask in English, not SQL."
The SQL Scribe: AI or Human?
But here's the rub: SQL isn't dead. Ironically, it's more critical than ever in the AI era. Semantic layers, vibe coding, and AI agents all depend on high-quality, high-fidelity SQL. Therefore, the question isn't if SQL is needed, but who writes it.
AI models don't magically understand your data. They need a precise and well-defined language to interact with it. That language is SQL; semantic layers act as a sophisticated translator, steering the creation of SQL queries needed to pull the right data.
This isn't the end of SQL; it's the end of manual SQL writing for most users. Data professionals, however, now face a new challenge. They must ensure the underlying SQL generated by AI is robust, secure, and scalable. They must also build the high-fidelity semantic models that guide the AI's SQL generation.
So, while analysts can now ask questions in English, a new breed of data architect or AI engineer ensures that the semantic layers and generated SQL behind the scenes are impeccable. They're the unsung heroes, ensuring the AI's answers are accurate, not just plausible.
The SQL scribe has changed, but the script remains essential.
This post is a part of a multi-part series about ideas shared in the 2025 Semantic Data Layer Summit. It explores topics like: Is data visualization dying, or is it just getting started? How can AI best leverage data to ensure its insights are accurate? What role do semantic layers play in the agentic applications of today and tomorrow?
I don’t think SQL is dying,I think it’s just moving backstage. For most business users, that’s probably a good thing. Nobody wants to spend half their day debugging joins when they just need a clear answer.
But someone still has to make sure the logic underneath is solid. If the semantic layer is off, the whole natural language setup becomes a guessing game. So we’re not removing SQL, we’re just shifting who interacts with it.
It kind of feels like the difference between driving a car and building one. Most people just want to drive, but you still need engineers under the hood making sure nothing catches fire.