iPadOS 27 Shortcut Builder Guide: Build and Test AI Automations
By the end of this guide, you'll know how to create a shortcut using the iPadOS 27 shortcut builder's new plain-English prompt tool, how to inspect what Apple Intelligence generates, and how to decide whether the result is safe to run unsupervised. That last part is where most coverage stops.
For years, building a Shortcut meant knowing which actions existed, how to connect them, and what variables to pass between steps a process that kept the feature largely in the hands of technically minded users, as MacRumors reported after the WWDC announcement this week. The new "Describe a Shortcut" feature replaces that blank canvas with a plain-text prompt field: describe what you want, and Apple Intelligence assembles the workflow. Critically, the generated shortcut lands inside the same editor that's always been there every action visible, every setting accessible rather than disappearing behind a chat interface, according to AppleInsider's hands-on testing published today.
That transparency is the feature's most important design decision. It's also the clearest signal that Apple isn't pitching this as a finished-product machine. It's an onboarding ramp and first-draft generator. For simple automations, it can get you to a working shortcut faster than manual assembly ever did. For complex ones, it gets you most of the way there and then needs you.
Scope note: This guide draws on hands-on demo reporting from WWDC 2026 and independent testing published this week. The public beta arrives this summer; the updated Shortcuts app ships with iPadOS 27 later this fall, per PCMag and TMC Insight. The workflow described here reflects pre-release behavior, and specific details may change before launch.
Before you start: hardware, software, and settings requirements
Not every iPad qualifies. Confirm this before reading further.
- Apple Intelligence requires an iPad with an M1 chip or newer. Older models won't run "Describe a Shortcut" regardless of software version. The same hardware gate applies to iPhone 15 Pro or later and any M1-or-newer Mac, per PCMag.
- Apple Intelligence is not available in all regions, even on supported hardware, as TMC Insight noted after the WWDC announcement. If you don't see the prompt field described below, check Settings → Apple Intelligence & Siri to confirm the feature is enabled and available in your locale.
- iPadOS 27 is currently in developer beta. The updated Shortcuts app ships with the OS no separate download required.
If your hardware qualifies and Apple Intelligence is active, you're ready.
How to create shortcuts in iPadOS 27: start with the right prompt
The UI change is minimal. The practical shift is large.
When you tap + to create a new shortcut in iPadOS 27, the traditional action browser gives way to an empty text field. What you type there determines everything. Apple Intelligence interprets the request and assembles the steps automatically, per AppleInsider you no longer need to know which actions exist before you start.
Vague prompts produce vague shortcuts. "Help me be more organized" gives the system nothing actionable. Concrete prompts with a defined trigger, a clear action, and a stated output consistently perform better. Use this pattern as a starting point:
When [trigger or condition], [action], and [output or destination].
Apple's own showcase example follows this structure precisely. A user describes sending their partner an ETA when they leave work, and Shortcuts chains a location trigger to an Apple Maps ETA calculation to a Messages send pulling together three separate system actions without the user needing to know any of them existed, per TechCrunch. The Apple Newsroom offers further examples: setting a morning alarm based on the next day's first Calendar event, or turning on porch lights when a delivery notification arrives.
A few prompt habits that consistently improve results:
- Name specific apps when the task touches more than one "using Calendar and Messages" rather than leaving it implicit
- State conditional logic explicitly: "if it's raining, do X; otherwise do Y" rather than implying it
- Break multi-part automations into separate shortcuts rather than prompting one shortcut to do everything; focused tasks generate more reliably than compound ones
- For text-manipulation tasks reformatting, stripping parameters, parsing input be especially precise. This category produced the most failures in hands-on testing
A note on complexity: Shortcuts involving multiple conditions, branching logic, or data passing between several apps are more likely to need corrections after generation, based on AppleInsider's testing. That's not a reason to avoid them. It's a reason to expect editing work, not a finished result.
Build your first test shortcut before tackling anything ambitious
Get comfortable with the generate-then-inspect workflow on something low-stakes first. A good starting point: a manual shortcut that enables Low Power Mode and starts a 25-minute timer. Two actions, no external apps, nothing that sends messages or touches files.
Here's how that looks in practice:
- Open the Shortcuts app and tap + to create a new shortcut. In the prompt field, type: "Enable Low Power Mode and start a 25-minute timer."
- Wait for Apple Intelligence to generate the workflow. For a two-action shortcut like this one, the result is typically correct and requires no editing as AppleInsider notes, simple tasks like this sit comfortably within what the tool handles well.
- Read every action before running it. Confirm Low Power Mode is toggled on, not off. Confirm the timer is set to 25 minutes, not some other value. Simple shortcuts still occasionally misread a detail.
- Run it manually. Watch it execute. Both actions should fire in sequence. If they do, you have a working shortcut and a concrete sense of what correct output looks like.
- Try a plain-language edit. Type something like "make the timer 30 minutes instead" in the prompt field and confirm the action updates in the editor. This is the fastest way to understand what the iterative workflow actually feels like.
The shortcut itself isn't the point. The habit is. Reading what the AI generated before trusting it matters far more when the shortcut sends a message to someone or modifies a file you can't easily undo.
Step 2: read what the AI actually built before you run anything
Apple Intelligence generates the first version of the workflow and places it in the standard Shortcuts editor, where every action remains visible and every setting is accessible, per AppleInsider. This is where the real work begins.
Inspect the output in this order:
- The trigger. Confirm the automation fires under the right conditions location, time, app state, or manual. If you intended to run this manually, verify no automatic trigger has been set.
- App permissions and inputs. Each action that touches a specific app, contact, or file should reference the correct item. AI-generated shortcuts sometimes leave placeholder variables or make assumptions about which contact, calendar, or account to use.
- The action sequence. Read through each step and confirm the logic matches your request. The system can select actions that are plausibly related but don't fully accomplish the task the structure looks right, but a step does something adjacent to what you need rather than exactly what you need, per AppleInsider.
- Output and destination. Verify where the shortcut sends its result a notification, a message, a file, the clipboard and that it matches what you intended.
- Whether it's safe to run unsupervised. If the shortcut sends messages, modifies files, or interacts with third-party services, run it manually several times before enabling any automation trigger.
To adjust the workflow after generation, describe the change in plain language in the same prompt field and Shortcuts will attempt to apply it, per Apple Newsroom. The "leaving work" example Apple demonstrated at WWDC shows this in practice: a user added "also start playing my podcast" after the initial build, and Shortcuts incorporated it, as TechCrunch reported. Described edits work reliably for simple additions. For structural changes to branching logic or multi-step sequences, editing actions directly in the editor is more predictable.
iPadOS 27 automations from plain English: a practical reliability guide
Reliability scales inversely with task complexity. Here's where the tool sits right now, based on this week's hands-on testing.
Works well use freely:
- Single-purpose tasks with a clear trigger and one or two actions (enabling Low Power Mode and starting a timer, for instance)
- Well-defined location or time triggers connected to a single app action
- Tasks that stay within one app ecosystem
Works as a first draft expect editing:
- Multi-step automations pulling from two or three apps
- Conditional logic where the AI gets the structure right but leaves fields or variables incomplete
- Calendar-based triggers with dynamic content
Requires significant manual work don't rely on AI alone:
- Text manipulation tasks, particularly anything involving parsing or reformatting input data
- Conditional branching with multiple outcomes
- Anything you plan to run on a schedule without reviewing it first
Two failure cases from this week's independent testing illustrate how wide that range is.
AppleInsider tested a shortcut to strip tracking parameters from URLs on the clipboard. The generated workflow looked structurally reasonable but failed repeatedly in practice. After multiple rounds of prompt revisions and manual adjustments, the model sometimes classified clipboard contents as a 404 page or denied that any URL was present at all. Even after manual tweaking forced the model to acknowledge a URL existed, it still didn't clean the tracking parameters.
PCMag tested a daily weather-and-outfit shortcut on iPhone. The shortcut ran and displayed current conditions, then returned an error stating that forecast data wasn't supported and directing the user to the Weather app and still recommended "consider light layers" before doing so.
Both cases reflect the same underlying pattern: the AI identifies the right general direction but either selects the wrong specific action, mishandles the input, or runs out of capability mid-execution. Worth knowing before you build something you're planning to trust.
What this feature is, and what it isn't yet
The best way to understand "Describe a Shortcut" is as an AI-powered onboarding layer, not a replacement for understanding how Shortcuts works. It solves the blank-canvas problem that kept automation inaccessible to most users, and it does that well. Every generated workflow surfaces in the standard editor, fully inspectable and editable Apple isn't turning Shortcuts into a chatbot, as AppleInsider noted, and that design choice keeps the user in control rather than hiding the work behind a conversational interface.
What it doesn't do yet is reliably generate complex, multi-condition workflows that run correctly without review. The hands-on evidence from this week is consistent on that point: simple tasks work, compound tasks need inspection, and text-manipulation tasks are genuinely unreliable in the current pre-release state.
The public beta arrives this summer. Start with one simple shortcut something you might have built manually before and use the AI to get a first draft. Read every action before running it. Build that habit now, while the stakes are low, before you're depending on any of this for something that matters.
The feature will improve. It already changes who can start building automations. Whether it changes how reliably they finish is the question the fall release will need to answer.
Comments
Be the first, drop a comment!