Zilliz predicts end of SQL dominance
Natural language interfaces powered by AI agents will become the primary method for querying databases in AI-driven workloads by 2026, according to Zilliz, displacing SQL to a secondary role for traditional analytics.
The prediction from the vector database vendor reflects growing limitations of relational database syntax for AI applications that increasingly operate on vector embeddings – semantic representations of text, images, audio and multimodal content.
"We're entering an era where talking to your database will be more productive than scripting against it," said James Luan, VP of Engineering at Zilliz. "SQL will still matter, but it no longer defines how people interact with data."
“SQL was never designed for similarity search across thousands of dimensions,” Luan added. “AI workloads require a semantic retrieval layer, not a relational one.”
Natural language interfaces allow users to describe requirements in plain language while AI agents automatically translate requests into execution plans. This approach reduces reliance on SQL specialists to translate business requirements into queries.
The shift addresses challenges organisations face managing complex AI workloads. Vector similarity searches across thousands of dimensions cannot be expressed efficiently in traditional relational syntax, according to the company.
Zilliz claims its purpose-built Milvus vector database delivers 60 per cent lower latency and 4.5 times higher throughput than PostgreSQL with pgvector extension under identical vector search conditions, based on internal benchmarks. Independent verification of these performance claims was not available.
The company states more than 10,000 organisations globally use Milvus and its managed Zilliz Cloud service for AI applications including semantic search, recommendation systems and retrieval-augmented generation. These deployments operate at billion-vector scale with sub-10 millisecond query latency, according to the vendor.
The prediction comes as enterprises increasingly deploy AI agents requiring realtime access to unstructured data. Vector databases store information as mathematical representations enabling semantic similarity searches rather than exact keyword matching.