Looker Data Apps is a new preview feature from Google’s Looker that lets business users generate production-grade, interactive web applications using plain-language prompts, while keeping all data access governed by Looker’s existing semantic layer. The post opens with a familiar pain point: dashboards offer fixed chart libraries and standardized grids, so any custom analytical experience — an executive briefing, a customer portal, or an interactive explainer — forces teams into building separate custom applications. AI tools make quick prototypes, but those prototypes stall because they lack governed data, row-level security, an infrastructure owner, and a clear security review path. Looker Data Apps closes that gap by having an agent generate a real application hosted natively inside Looker.

The post defines a data app as a bespoke, interactive web application generated by an AI agent and hosted within Looker. Unlike dashboards, which provide top-down visibility, data apps are active systems of action built around a specific operational workflow. They go beyond grid tiles with pixel-perfect layouts for a particular role, let business users input parameters, test scenarios, run live simulations, and update underlying records, and connect analysis to execution tasks such as updating source records or dispatching events to tools like Salesforce and Jira. Users can build these interfaces with AI developer tools like Claude Code or Antigravity CLI, grounded directly by Looker’s semantic layer.

Data apps are positioned as an expansion of the semantic layer, not a replacement for dashboards. Dashboards remain ideal for aggregate visibility and passive reporting, while data apps handle active execution by letting users modify parameters and trigger downstream tasks. Dashboards offer consistency through standardized chart libraries; data apps offer flexibility through custom web components and specialized interactions. All logic stays governed centrally via the LookML semantic layer, so metric discrepancies and hallucinations are eliminated whether users view a KPI dashboard or adjust budgets inside a data app.

The enterprise-oriented design emphasizes governance and operational efficiency. Every measure is grounded in the semantic layer, and every app inherits existing semantic models, row-level security, and access controls. Built-in query inspection lets teams audit every line of generated SQL to eliminate metric hallucinations. Deployment is zero-code by default: agent-generated React applications run directly inside Looker with no code, and teams can also use LLMs like Claude Code and Antigravity CLI to build interactive apps and deploy them as data apps in minutes. A key cost property is that there are no runtime token costs: LLM tokens are consumed only during design or generation, while end-user interactions run on native Looker APIs and models. That makes the marginal cost of serving 50 or 5,000 users effectively flat. Data apps can also be embedded into external portals using the same workflows used for dashboard embedding, with no new pipelines, no custom application code, and no second permission model. The article concludes that the trade-off between basic dashboards and expensive bespoke systems is ending, and Looker Data Apps is now available in preview.