Apple opened Siri AI integrations to all developers on November 3, 2026, through a Developer Preview in the iOS 27.3, iPadOS 27.3, macOS 27.3 and visionOS 27.3 betas. Apps can expose content and custom actions through App Intents, while a new skills feature lets developers combine several actions into one experience, according to Apple’s announcement.
The preview also comes with consequential access conditions. Siri AI requires an Apple Intelligence-enabled device, excludes users under 13 and remains unavailable in the EU on three platforms. Apple also disclosed daily limits for certain server-side AI features, with expanded access planned for a fee.
For developers, the opportunity is to make app functions accessible through conversation. The planning challenge is that availability, usage policies and future AI pricing remain only partly defined.
App Intents Connects App Actions to Siri AI
App Intents is the integration path Apple names for bringing an app’s content and actions into Siri AI. Developers supply the functionality, such as checking a delivery, booking a reservation or placing an order. Siri provides the conversational interface through which people request those actions.
Apple says integrations can use capabilities including:
- Personal context understanding, helping Siri interpret requests in relation to the user.
- Onscreen awareness, allowing requests to relate to what is displayed.
- Natural-language understanding, letting people ask for an action in their own words.
The new skills feature addresses more complex requests by letting developers teach Siri how to combine several actions into a single experience. Apple also says Siri AI can surface any entity in the Spotlight semantic index, expanding beyond entities that fit an existing schema. These are capabilities described in the developer announcement, not independently measured reliability results.
The useful distinction is between exposing one app action and designing a complete workflow. A delivery-status lookup has a narrow outcome. A skill combining several actions introduces more points where missing information or an unavailable service could interrupt progress.
I’d start with bounded actions whose results are easy to verify before building longer skills. That recommendation follows from the workflow structure: validating each component makes it easier to diagnose a failure when several components are combined. It is not a claim that Apple’s preview has demonstrated a particular failure rate.
