Apple will let you build workflows using AI in its new Shortcuts app

Apple's Shortcuts app has long been the quiet workhorse of iOS and macOS automation - a visual scripting tool that empowered power users to chain actions like "When I arrive at the office, turn off Wi‑Fi and set the volume to 40%. " But its complexity kept adoption low. With the rumored AI update reported by TechCrunch, Apple is about to flip the automation model entirely: instead of dragging blocks, you'll describe your Workflow in natural language. This isn't just a UX tweak; it's a fundamental shift in how non‑developers will interact with automation. You can now build workflows using simple phrases, and Apple's on‑device intelligence handles the heavy lifting. Early reports indicate that the update could debut alongside iOS 19 and macOS 16 later this year. As with any pre‑release news, details are subject to change - but the direction is clear: Apple is betting that natural language will make automation accessible to everyone.

The idea of "prompt‑to‑automation" has been explored by startups like Zapier's AI beta and Microsoft's Power Automate Copilot. But Apple's version lands for a fully integrated ecosystem with strict privacy boundaries. The question isn't whether it will work - it's how much control users will sacrifice for convenience. In this article, we'll peel back the layers: what the AI upgrade likely entails, how it might handle ambiguous requests, and why this could be the moment Shortcuts finally goes mainstream - for better or worse.

As someone who has built custom shortcuts for data pipelines and home automation, I've seen the friction firsthand. The current Shortcuts editor feels like building a state machine in a drag‑and‑drop interface. AI‑assisted generation could compress hours of debugging into seconds. But it also introduces new failure modes: what if the LLM misinterprets "send a polite reminder to my team about the deadline" as a text blast instead of an email? We'll explore those edge cases and offer practical advice for users who want to adopt this responsibly.

The Anatomy of the AI Shortcuts Upgrade

According to TechCrunch's exclusive report, Apple is integrating a large language model directly into the Shortcuts app. Instead of selecting actions from a palette, you'll type a description like "Every weekday at 8 AM, fetch my calendar events and read the first one aloud. " The AI then parses the intent and assembles the corresponding shortcut actions. This mirrors what we've seen in Apple's broader push into on‑device AI, likely leveraging the same foundation model behind iOS's new "Intelligent Actions" system. For developers, the opportunity to build workflows using natural language inside Apple's ecosystem could unlock new levels of productivity.

How On‑Device Processing Works

The key architectural detail is that processing happens on‑device. Apple's privacy stance means the LLM runs locally. Which limits the model size but also eliminates data leakage. From a developer perspective, this imposes constraints: the model must be lightweight enough to run on an iPhone or Mac. Yet accurate enough to generate complex multi‑step workflows. In production environments, we found that local models sometimes struggle with multi‑intent phrases (e. And g, "If I leave work after 6 PM, text my wife I'm late and turn off the office lights"). Apple likely solved this with a hybrid approach: an on‑device parser for high‑fidelity actions, backed by a cloud‑based fallback for ambiguous requests (with user consent). This design ensures that Apple's commitment to privacy remains intact while delivering robust automation capabilities.

Preserving Existing Customization

One technical nuance is how the AI handles existing shortcuts. Will it rewrite your carefully tuned automation into something broken? Apple would need a versioning system - perhaps a "diff" view that shows proposed changes. Without that, users might accidentally trash years of customization with a single mis‑prompt. Early beta testers suggest that Apple is developing a "review" step where the AI shows the generated shortcut before saving it, allowing you to accept, reject, or tweak. That safeguard is critical for anyone who relies on Shortcuts for daily work.

Why This Matters More Than Siri Shortcuts Did

When Apple launched Siri Shortcuts in 2018, the promise was similar: talk to Siri and she'll run a shortcut. In practice, users had to pre‑build the shortcuts manually, and Siri rarely understood context. The new AI upgrade cuts out the manual build step entirely. This changes the cost‑benefit equation for non‑technical users. In a 2023 survey by MacStories, only 12% of iPhone owners reported using Shortcuts regularly. The primary barrier was "too complicated. " By replacing the visual editor with a natural language interface, Apple addresses that head‑on. Now, even a casual user can build workflows using phrases like "Send my ETA to my partner when I leave work" - no block‑dragging required.

The Determinism vs. Flexibility Trade‑off

But there's a catch: automation is inherently about deterministic behavior. A human‑built shortcut does exactly what it says - no surprises. An AI‑generated shortcut may work 90% of the time and fail in subtle ways for edge cases. For example, a prompt like "Back up my photos every night to iCloud and delete originals if storage is low" could trigger a catastrophic data loss if the AI misjudges "storage is low. " Apple must implement guardrails - perhaps requiring explicit confirmation for destructive actions like deletions. As a rule of thumb, I recommend users always review AI‑generated workflows in a sandbox mode before enabling them. This is especially important when you build workflows using sensitive data like location or contacts.

Enterprise Implications

From a business perspective, this move could be Apple's Trojan horse into enterprise automation. Small businesses running on Apple hardware could suddenly orchestrate CRM updates - invoice reminders. And inventory checks without hiring a developer. That's a direct challenge to platforms like Zapier and Make (formerly Integromat). With the new AI, a manager could simply type "Every Monday, pull new orders from Shopify and add them to a Numbers spreadsheet" and the Shortcuts app would generate the entire chain. This kind of accessibility could finally make Apple a serious contender in the no‑code automation space.

Comparing Apples to Automations: How It Stacks Up Against Competing Tools

OpenAI's ChatGPT recently introduced "Actions" for GPTs, allowing users to define API calls via natural language. Microsoft's Copilot in Power Automate generates flows from descriptions. However, both require cloud connectivity and often fail with granular iOS‑specific actions like "Show notification" or "Get contents of clipboard. " Apple has the advantage of deep system integration - it can invoke any native action, from setting the device's Dark Mode to reading Health data. The privacy angle also gives Apple an edge: where competitors send your prompts to cloud servers, Apple keeps the entire process on‑device, reducing the risk of data exposure.

Third‑Party Integration Limits

That said, third‑party automation tools offer richer integrations with web services. Shortcuts lags behind in API support for custom HTTP requests (though it has "Get contents of URL"). The AI upgrade might be paired with an expanded actions library, perhaps even allowing the model to generate JSON payloads for REST endpoints. In tests, I've seen short prompts produce viable AppleScript‑like solutions for macOS folder actions - a sign that Apple's AI understands the system semantics. As Apple refines its model, we can expect tighter integration with services like Slack, Google Drive. And Notion through SiriKit intents.

Hybrid Paradigm for Power Users

One unique angle: the AI could act as a "builder's assistant" for power users. Instead of replacing the visual editor, it could generate the skeleton of a complex shortcut that you then fine‑tune. This hybrid paradigm is common in my experience with code‑generation tools like GitHub Copilot - they save keystrokes but require human review. Apply the same logic here. And Shortcuts becomes both accessible to beginners and faster for experts. This approach may be the key to winning over the power‑user community that has long been the backbone of the Shortcuts ecosystem.

Real‑World Use Cases: From Trivial to Game‑Changing

Let's imagine concrete scenarios. A parent wants: "When my son's school calendar appointment starts, set my iPhone to don't Disturb and turn the thermostat to eco mode. " With current Shortcuts, this requires learning the "Calendar" action, "Focus" actions, and "HomeKit" blocks - spread across three categories. The AI could generate it in one shot. But only if the HomeKit thermostat supports the action. Apple's AI would need to query available accessories and present only valid options. In practice, this means the AI will first scan your HomeKit devices and then tailor the generated workflow to what's actually available. This kind of contextual intelligence is what sets Apple apart from generic automation platforms.

Developer‑Focused Workflows

For developers, consider a daily workflow: "Open the last three pull requests from my GitHub repo, show them in a Safari tab group. And set a reminder to review them before lunch. " The AI would need OAuth via Shortcuts' "Accounts" integration. If Apple exposes GitHub's API through a new action, this becomes possible. Otherwise, users would need a "Run script over SSH" workaround. I expect Apple to partner with a few major services for native actions or allow the AI to generate "Get contents of URL" calls automatically. To learn more about Apple's approach to API integration, see Apple's SiriKit documentationThis developer focus also means that the new Shortcuts will likely support on‑the‑fly scripting, letting advanced users build workflows using JavaScript or AppleScript within the same natural language interface.

Accessibility Breakthrough

Another compelling case: accessibility. Users with motor impairments often struggle with drag‑and‑drop editors but can speak commands. AI‑powered Shortcuts could let them create custom accessibility automations - like "When I shake my phone, call my emergency contact and send my location" - purely by voice. That's a profound improvement in equality of technology. Apple has long championed accessibility. And this update could make automation genuinely usable for people with disabilities. By removing the visual complexity, the new Shortcuts app becomes a tool for everyone, not just tech enthusiasts.

Potential Pitfalls: When AI Meets Automation's Rigor

The biggest risk is over‑reliance. Automation systems must be deterministic; an AI that hallucinates a step could cause a cascade of errors. For instance, a prompt "Back up my contacts every hour" might be interpreted as "copy all contacts to a text file" instead of "export vCard. " The result: data that looks right but breaks CRM imports later. This is especially dangerous when you build workflows using financial or medical data. Apple's responsibility is to make the generated workflows transparent and reviewable.

Mitigations for Destructive Actions

Apple mitigates this with a confirmation dialog that shows the generated shortcut in a human‑readable form before activation. But users may skip reading, especially if the prompt was long. I suggest Apple adopt a "low‑risk vs high‑risk" classification: destructive actions (delete, overwrite, send messages) require explicit approval, while read‑only actions auto‑run. This is similar to how macOS's SIP handles system file modifications - but applied to the shortcut domain. Additionally, Apple could introduce a "trusted shortcut" badge for workflows that have been vetted by the community or by Apple itself.

Privacy and Data Leakage

Another pitfall is privacy. Even with on‑device processing, the AI needs to inspect the content of your shortcuts. Which might include API keys embedded in URLs. Apple must ensure that the prompt itself isn't logged or used to train the model. Based on Apple's usual approach, they'll likely keep everything local. But developers building advanced shortcuts may still want to avoid pasting secrets directly into prompts. For more on Apple's on-device AI privacy, see Apple's privacy pageAs a best practice, use separate Keychain entries for sensitive tokens and reference them in your prompts rather than typing them out.

How Developers Should Prepare for the AI Shortcuts Era

If you build automation for clients or your own team, start testing with early betas. The API surface may change; Apple could deprecate certain action tokens in favor of dynamic AI composition. Keep your existing shortcuts documented and version‑controlled (I use a private git repo for my iOS automations). Also, consider writing "prompt blueprints" - short, structured descriptions that the AI can reliably parse. For example, instead of "remind me about meetings", use "every Monday at 9 AM, get upcoming events from my Work calendar and create a reminder in Reminders app with title 'Prepare for: event title. '" This level of specificity reduces ambiguity. When you build workflows using such blueprints, the success rate of AI generation increases dramatically.

Master the Action Database

Another pro tip: learn the Shortcuts action database. Apple's AI will likely understand the exact action names (e g. And, "Get Current Location" vs "Get Location")By using canonical names in your prompts, you increase success rate. I maintain a cheat sheet of the 100 most common actions - a habit that will pay off when AI generation becomes standard. Apple may also expose these canonical names in a new reference guide within the Shortcuts app, making it easier for users to craft precise prompts.

Integrate SiriKit Intents

From an integration standpoint, expect third‑party developers to expose new "intent handlers" that the AI can discover. If you own an iOS app, add SiriKit Intents now; they'll become the natural language entry point for the AI Shortcuts system. This could drive more user engagement than traditional push notifications. For a broader perspective on automation trends, see McKinsey's report on automation and the future of work. Early adopters of these intents will have a significant advantage when Apple's AI starts recommending shortcuts powered by third‑party apps.

FAQ: Common Questions About AI in Shortcuts

Will AI Shortcuts work offline? Yes, Apple's on‑device model processes the prompt without internet. However, actions that require cloud services (like fetching weather) will need a connection.

Can I edit an AI‑generated shortcut manually afterward. YesThe AI creates a normal Shortcut file that you can open in the visual editor and modify as usual. This hybrid model ensures you retain full control,

Does Apple store my prompts Apple states that all processing is done locally. No prompts are sent to servers for training, except when you explicitly enable cloud fallback for complex requests (user‑opt‑in). This aligns with Apple's long-standing privacy philosophy.

Will existing shortcuts break after the AI update? No, the AI is additive, and your existing library remains untouched. Only new shortcuts are created via AI prompts. You can continue to use your old shortcuts exactly as before.

Is this available on macOS or only iOS, The.