Google launches Looker Data Apps to build secure tools

Google has launched Looker Data Apps in preview, adding an application-building layer to its Looker analytics platform.

The offering is designed to let users create interactive web applications inside Looker using natural-language prompts or AI coding tools, while relying on the same semantic models, access controls and row-level security already used across the platform.

It is intended to address what Google sees as a gap between dashboard software and custom-built applications. In the company's view, dashboards often limit users to fixed layouts and chart libraries, while bespoke operational tools usually require separate engineering work, infrastructure ownership and security review before wider rollout.

Looker Data Apps hosts generated applications natively within Looker. Google describes them as multi-page interfaces that can support operational workflows, user inputs, scenario testing, simulations and updates to underlying records.

That puts the product in a different category from standard dashboards, which are generally used for reporting and top-level visibility. Data apps, by contrast, are meant to connect analysis with operational tasks, including changes to source records or actions involving other business systems such as Salesforce and Jira.

Governed data

A central part of the launch is the use of Looker's semantic layer. Every application built with the new product inherits existing semantic models, security settings and access controls, with calculations kept consistent through the same LookML framework used elsewhere in Looker.

Teams can also inspect generated SQL queries, which Google presents as a way to audit how metrics are produced. The feature is aimed at organisations that want to use AI-assisted development tools without creating a separate governance structure for applications built outside their established analytics environment.

Users can create applications without writing code by deploying agent-generated React applications directly inside Looker. Teams can also use external AI developer tools, including Claude Code and Antigravity CLI, to design interfaces and then deploy them into the platform.

Production path

Google framed the launch as an attempt to move AI-generated prototypes into live business use. It argued that many rapid application demos fail to progress because they lack access to governed data, a clear infrastructure owner or a route through internal security checks.

By keeping the apps inside Looker, organisations can avoid introducing separate permission models, custom application code or new deployment pipelines, according to Google. The same embedding workflows used for dashboards can also place the new applications into external portals.

Google also highlighted operating cost. Large language model tokens are used during the design or generation stage, but end-user interactions then run on Looker APIs and models rather than continuing to consume AI tokens at runtime.

That pricing and architecture may appeal to enterprises weighing whether to expose AI-built tools to large user groups. Google said the marginal cost does not change whether an application is used by dozens or thousands of people.

Broader shift

The launch reflects a wider shift in business intelligence software, as vendors try to extend analytics tools beyond reporting and into workflow execution. Rather than limiting users to dashboards, software groups are increasingly using AI to generate interfaces that let staff test scenarios, enter data and take action without leaving the analytics environment.

For Google, the move also strengthens Looker's position as more than a dashboarding product. By adding app generation tied to its semantic layer, the company is seeking to make the platform part of day-to-day operational processes as well as reporting.

Looker Data Apps is available in preview.