Apple Intelligence in Practice

Apple Intelligence is a layered platform. Pick the wrong layer and your AI feature ships broken.

That's the painful lesson developers learn in their first six months of Apple Intelligence work. They reach for the on-device language model when a system feature would do. They build custom rewriting UIs that fight Writing Tools. They treat App Intents as legacy plumbing, then wonder why Siri can't find their app.

This book is the map.

Apple Intelligence in Practice is a working developer's guide to shipping production-quality AI features on iOS, iPadOS, and macOS written for Swift developers who have already shipped at least one app and want to integrate Apple Intelligence properly. Not chase the marketing. Not reinvent system features. Not get pulled from App Store review for sloppy implementation.

What you'll learn:


The decision framework for choosing the right layer of Apple Intelligence for each feature the architectural discipline that separates clean implementations from months of rewrites
How to build production integrations with the Foundation Models framework on-device, Private Cloud Compute, and third-party models like Claude and Gemini through the same unified Language Model protocol
How to generate reliable structured output with @Generable and @Guide, eliminating the parsing bugs that plague early LLM apps
How to let the model invoke your code through the Tool protocol, with the description patterns that make tool selection actually work
How to make your app reachable to Siri, Shortcuts, Spotlight, and Apple Intelligence orchestration through assistant schemas and App Intents
How to build semantic search using NLContextualEmbedding and the 2026 Spotlight-powered search primitive without misusing the language model
How to ship cleanly across the capability matrix with graceful degradation, voice-first design, and App Store review compliance


This is not a beginner's book. It assumes you can read Swift, understand SwiftUI bindings, and have shipped real apps before. It will not teach you closures, optionals, or @MainActor.

This is also not a documentation rewrite. Apple's documentation is good, online, and updated more often than any book can match. Instead, this book teaches you how to think about the platform the decision framework, the architectural patterns, and the production discipline that survive whatever Apple ships in the next SDK.

Twelve chapters. Five reference appendices. Built on the post-WWDC 2026 state of Apple Intelligence: AFM 3, multi-mode Foundation Models, agentic primitives, the Spotlight semantic search tool, and the third-party Language Model protocol that lets Claude and Gemini plug into the same API as Apple's on-device model.

For working Swift developers who need to ship Apple Intelligence features on a deadline, this is the book that gets you there cleanly.