Building a complete Drupal site demands expertise across content types, taxonomy, menus, block configuration, and styling. What if a non-technical user could describe what they need in plain language and get back fully structured, styled pages, then refine everything through conversation? This session presents a Drupal 11 architecture that turns page building into a conversational AI experience, empowering non-technical users to create and iterate on complete sites without writing code or navigating complex admin interfaces.

Prerequisite

Familiarity with Drupal's editorial interface and content architecture (content types, blocks, menus, taxonomy). A general awareness of how LLM tool-calling works is helpful but not required: the session explains the relevant concepts as they arise.

Target Audience

A Drupal developer or technical architect who builds or maintains component-driven sites (Layout Builder, Drupal Canvas, or similar) and wants to understand how to layer AI-assisted authoring on top of existing page-building infrastructure.

Outline

The session opens with a ten-minute live demo: building a complete site from a blank installation using natural-language prompts, homepage, about page, taxonomy-driven blog listing, navigation menu, styled footer, and AI-generated images, then refining it through contextual chat. The demo concludes by showing all changes staged inside a Workspaces Extra workspace for editorial review before publishing.

The tooling segment explains the discovery-first architecture: function-call tools that let the AI introspect component schemas, layouts, design tokens, and site structure, adapting automatically to any page-building approach without hardcoded lists.

The workflow segment covers creation and editing: AI-proposed structures validated against Drupal definitions, a self-correction loop for automatic error repair, context-aware editing of existing pages, asynchronous image generation, and persistent conversational memory across sessions.

The styling segment details the per-component CSS cascade across four levels (global tokens, component defaults, per-page overrides, and per-instance styles) and how design-system introspection keeps AI output on-brand.

The final segment discusses provider portability via the contrib AI module, editorial governance through Workspaces Extra, honest limitations around prompt sensitivity and generation cost, and where human editors still do better.

Learning Objectives

Attendees will leave with a concrete understanding of what AI-assisted site building can deliver in Drupal today, including: the architectural patterns behind a reliable conversational page-building system (validation, self-correction, discovery-first tooling); how to design a per-component styling cascade that keeps AI output within a site's design system; how Workspaces Extra provides editorial governance for AI-generated content; how the same approach adapts across page builders (Layout Builder, Drupal Canvas, and others); and a realistic assessment of when this approach adds value versus when traditional editing is the better choice.

Experience level
Intermediate