Google Cloud has launched QueryData, a tool that translates natural language into database queries with claimed near-100% accuracy, positioning it as a more reliable alternative to direct LLM-generated queries in AI agent workflows. It requires teams to define upfront 'context' — schema descriptions, relationships, and deterministic instructions — before integrating into agent-driven pipelines. QueryData supports AlloyDB, CloudSQL (MySQL/PostgreSQL), and Spanner, and can be used within BigQuery's data agents or via API in custom multi-agent systems. Analysts note it reduces prompt engineering but shifts complexity to upfront design and ongoing schema maintenance, creating a new 'data access engineering' workload. Strategically, it reflects Google's focus on the data layer, contrasting with OpenAI's API approach, AWS connectors, and Microsoft's Copilot integrations. The tool is currently in preview.
Sort: