Google's QueryData Launch

Google Cloud has put a new layer between artificial intelligence agents and company databases: QueryData, a preview tool that turns plain-English requests into executable queries. (cloud.google.com) The launch was announced by Google Cloud on April 10, 2026, and the company said QueryData works with AlloyDB, Cloud SQL for MySQL and PostgreSQL, and Spanner for GoogleSQL. Google described it as a way to build “agentic experiences” that query operational databases directly. (cloud.google.com) In practice, the product splits one job into two parts: a model interprets the user’s intent, then QueryData checks that intent against database-specific context before producing a query. Google’s documentation says customers have to supply that context, including templates, facets, and business rules about how data should be accessed. (docs.cloud.google.com) That setup targets a common failure in artificial intelligence database tools: large language models can write SQL syntax, but they do not know a company’s table names, joins, or internal definitions unless those rules are provided. Google said the context layer is the code that guides the system toward correct answers for a specific database. (cloud.google.com) Google is pitching reliability as the selling point. In its launch post, the company said QueryData delivers “near-100% accuracy,” tying the product to Google Cloud’s performance on the BiRD benchmark, a widely used test for natural-language-to-database systems. (cloud.google.com) The company also framed QueryData as a narrow tool, not a general chatbot for data work. InfoWorld reported that Google separated the fuzzy reasoning done by models from the brittle work of generating safe, valid SQL that can actually run against production systems. (infoworld.com) That matters for companies using databases that run orders, inventory, billing, or support systems, where a wrong query can return the wrong records or fail outright. Google’s examples include shopping assistants, field operations tools, and other applications that need live answers from operational data, not just summaries from a dashboard. (docs.cloud.google.com) QueryData also fits into a broader Google Cloud push to add Gemini-powered features across its data products. At Google Cloud Next 2025, the company highlighted artificial intelligence features in databases, analytics, and governance tools, including natural-language interfaces for data work. (cloud.withgoogle.com, techcrunch.com) Google said Hughes Network Systems has already deployed QueryData in production, offering an early customer example as the product enters preview. The immediate test now is whether enterprises will do the labor of defining enough database context to make those plain-English requests dependable at scale. (cloud.google.com, infoworld.com)