Power Apps can now build a full model-driven app from a prompt

Microsoft has published its September 2026 feature update for Power Platform. The biggest change for makers: you can now describe a model-driven app in plain language and have Power Apps generate it, and that capability is now generally available.

The update was posted on September 17 by Tiffany Treacy, Vice President of Product for Power Platform.

Power Apps

Build a model-driven app from a prompt (GA)

The new app-builder skill turns a natural-language description into a working model-driven app. It generates the tables, forms, views, charts and security roles for you. This is Microsoft's answer to the "describe it and get an app" builders that have spread across no-code tools this year.

AI help inside canvas apps (GA)

The canvas authoring agent plugin is also generally available, so makers can use agent-driven workflows while building canvas apps.

Generative pages grow up

Generative pages can now be embedded directly in model-driven app forms. In preview, they can also pull data from Power Platform connectors.

Smaller, welcome changes

• Six Fluent screen templates for canvas apps, based on Fluent 2 patterns. They give you responsive layouts and editable controls as a starting point.
• Bulk update for modern controls. Eligible controls can be updated together instead of one at a time. There are also new progress bar, avatar and spinner controls.
• A refreshed look for model-driven apps. The header and navigation refresh is now generally available. A display density setting, with three comfort levels, is in preview.

Power Automate

• Quick start cards on the home page give you direct entry points into common automation scenarios.
• Server-side search finds flows across an entire environment. Previously it only searched the flows loaded in your browser, which matters in large tenants.

Learning

Microsoft Learn has new training paths covering AI code generation, model-driven apps, canvas apps and the data platform.

What it means

Every major low-code platform now offers a way to go from prompt to working app. Microsoft's version stands out because it outputs a governed model-driven app, with security roles included, rather than just screens. For teams already on Dataverse, that makes AI-generated apps much easier to take into production.