Apple will let you build workflows using AI in its new Shortcuts app(techcrunch.com)

Shortcuts always felt like the cousin nobody invited to the family photo, and bolting an LLM onto it might finally give the ops team at my company a reason to stop building Zapier flows that cost us $40k a year. Curious whether the on-device model handles branching logic or just the easy summarize-and-route stuff.

Agreed, the productivity numbers in these demos never survive contact with real workflows. We piloted Shortcuts with AI actions across a 12-person mobile team and the time saved was eaten by debugging non-deterministic outputs in steps that used to be one-line scripts.

Tried wiring Shortcuts into our release checklist last week and it shaved maybe 4 minutes off a 45 minute process before the model hallucinated a build number that didn't exist. The 9% time savings evaporated the moment a junior shipped the bad number to TestFlight and we spent an hour rolling back.

Does the AI build steps from a plain-language prompt, or just suggest actions inside the existing visual editor?

Shortcuts has technically supported automation since 2018, and the dropoff rate on my team tells the story: of 11 devs I onboarded to a custom Shortcuts pipeline last year, 2 still use it because debugging a failed action gives you no stack trace, just a red box. Bolting an LLM onto that means I now get to guess why the AI step picked the wrong variable instead of why my hardcoded one did.

Eighteen months of using these for document review and the pattern that holds is that visual builders make the easy automations easier and the brittle ones break in quieter ways. A Shortcut that misroutes a privilege-flagged email fails the same way a bad macro did, except now nobody can read the logic to figure out why.

Shortcuts has had a non-AI automation builder since 2018 and the reason it never caught on outside power users wasn't the lack of a model, it was the broken triggers and apps that silently refuse to expose their actions. Dropping an LLM on top of that doesn't fix a Calendar action that still can't reliably create a recurring event. You're adding nondeterminism to a system people picked precisely because it ran the same way every time.

Workflows-as-code only works when the workflow is stable, and the whole reason people reach for Shortcuts is the long tail of one-off glue that changes every week. Watched a teammate spend 40 minutes coaxing a "rename these files and upload" shortcut into existence last year, then never run it twice; an LLM guessing the steps does not fix that the work was a one-off.

Shortcuts always broke the moment a step returned something unexpected, so an AI layer that adapts mid-flow is the missing piece. I wired a Siri Shortcut to a Claude agent last month and it finally handled "file this receipt" without me pre-defining every folder name and edge case.

Building the same three-step "transcribe, summarize, route to Slack" flow by hand with API calls and a cron job took me a weekend. If Shortcuts makes that drag-and-drop for non-engineers, half my internal glue scripts stop being worth maintaining.

Automating half a team means someone still owns the exceptions that AI kicks back. We route those through Zendesk macros, and the refund-over-$50 queue still needs a human before anything ships.

Everyone forgets Shortcuts already had non-AI actions for years and almost nobody chained them past a two-step "set a timer" flow. Dropping an LLM into that same drag-and-drop canvas doesn't fix the actual problem, which is that the moment a workflow needs a branch or a retry you're back writing real code. AI in Shortcuts makes the demo prettier and the ceiling exactly as low as it was.

Shortcuts has had a visual workflow builder since 2018 and most people still never touched it because stringing together actions meant learning Apple's mental model of variables and magic variables. Bolting AI onto that authoring step does nothing if the underlying actions still break every time an app updates its intents. The bottleneck was never writing the workflow, it was trusting it would run tomorrow.

Half my agents' tickets now resolve without a human, and the thing that made it work was not the visual builder in Zapier or Shortcuts, it was the fifty edge cases we wrote conditions for after watching them fail. AI-generated workflows will nail the happy path and leave you to discover the other 40 percent in production, one angry customer at a time.

Shortcuts still crashes if I rename a variable, but sure, let it write my workflows now.

Hand-rolling GPT calls with a Zapier duct-tape layer, and now Apple ships it as a drag-and-drop node.

We used to have 6 people on live chat and a queue that hit 40 tickets by noon. I wired up a Zendesk macro trigger plus a small model that drafts replies for password resets and refund status, and now two people handle what six did, mostly just approving drafts. The thing that surprised me was tagging: auto-categorizing intake cut our misroute rate from about 18% down to 3, which mattered more than the reply speed. If Shortcuts makes that kind of glue buildable without a dev sitting next to me, that's the part I'd watch.

Shortcuts has technically let people chain actions since 2018, and the graveyard of half-built automations proves the bottleneck was never the syntax. Dropping a language model on top does nothing about the real wall: an app has to expose an action for Shortcuts to touch it, and most of the ones I rely on daily still expose nothing.