professorkamrankhan@gmail.com 
While other tech companies rush to dominate artificial intelligence with bold promises and uneven results, Apple is building something slower, quieter, and more durable.
Out of curiosity I tested Apple Intelligence using the new Shortcuts action in iOS 26. Apple mentioned this feature at WWDC 2025, but it didn’t get much attention during the keynote.
The Shortcuts action gives users direct access to Apple’s large language model. You can choose the smaller on-device version or the more capable cloud model running in Apple’s private compute environment.
In other words, it lets you talk to Apple’s AI like you would with ChatGPT or Claude. Craig Federighi, Apple’s senior vice president of Software Engineering, and Greg Joswiak, senior vice president of Worldwide Marketing, have said Apple isn’t trying to build a chatbot.
But the new model action in Shortcuts works like one, at least sometimes. It gives you direct access to the language model without going through Siri. Depending on the prompt, it behaves like a simple, stripped-down chatbot.
In my testing, Apple Intelligence did well at basic alignment. Alignment refers to an AI system’s ability to follow human ethical norms and safety guidelines. That includes things like refusing harmful, deceptive, or illegal requests.
In one example, it refused to generate a fake doctor’s note. But it gave an answer when I wrapped it in a fictional scenario. However, that kind of response to trickery isn’t limited to Apple’s AI.
Large language models don’t work like traditional software. They don’t retrieve known answers from a database.
The AI companies feed them with human data, and the LLMs generate plausible responses based on training patterns. They can be persuasive and wrong at the same time.
Apple can’t avoid this problem because it’s a side effect of the technology. But the company can control how the system is used, where it runs, and what kind of mistakes users are most likely to see.
Why Apple is moving slowly
Companies and shareholders are clamoring for a piece of the generative AI action. They have to prove to Wall Street that they’re still making the numbers go up.
I suspect that’s the reason why Apple announced its AI features before they were truly ready. There are plenty of arguments about what went wrong but the ultimate answer is “everyone else is doing it.”
Testing Apple Intelligence in Shortcuts
Siri was never meant to be an AI. It was a voice interface for a rule-based system. If it couldn’t parse your request, it would kick you out to Safari.
That fallback gave Apple cover. If the results were bad, it was the internet’s fault and not that of Siri. That model worked when expectations were lower.
Apple Intelligence is a step into something far more unpredictable. Instead of prewritten responses, the new system uses a generative model trained on vast amounts of text. That gives it flexibility and risk.
It’s harder to sandbox, harder to test, and near-impossible to guarantee consistent answers. For a company like Apple, which prizes control and predictability, that’s a big philosophical leap.
Apple is betting on a horse
Right now, Apple Intelligence rewrites emails, summarizes notifications, pulls context from documents, and ties into apps like Calendar, Safari, and Notes. These aren’t show-stopping features.
But they’re useful. And they fit Apple’s broader pattern of releasing technology only when it improves everyday tasks. Apple Intelligence isn’t meant to be flashy, it’s meant to fade into the background until you need it.
Apple is skipping the first generation of consumer AI entirely. Chatbots are “AI 1.0.” These are models that talk back, sound smart, and try to answer anything.
But that form of AI is messy, hard to trust, and expensive to run. Apple is betting that users don’t actually want a chatbot.
They want help.
Testing Apple Intelligence in Shortcuts
Agentic AI, or “AI 2.0,” is the next goal in the generative AI space. Agents are tools that take action rather than hold a conversation. That’s exactly what Apple “demoed” at WWDC 2024.
That’s when AI begins to act more like a feature than a product.
Apple also has two strategic advantages that few rivals can match. First, it has brand loyalty that borders on cultural. Millions of users trust Apple to get things right eventually, and that gives the company space to move slowly.
Second, it has cash reserves large enough to outwait any hype cycle. Apple doesn’t need to win the first round of a new tech race. It just needs to be standing when the dust settles.
The illusion of a booming AI market
Some analysts insist Apple is behind in AI, but behind what, exactly? A race to launch half-baked chatbots with no business model?
Apple isn’t chasing engagement metrics or demo hype. It’s building tools people will actually use. If that looks slow from the outside, it probably means Apple is paying attention.
Apple is always late. The iPhone wasn’t the first smartphone. The iPad wasn’t the first tablet.
The Apple Watch entered the wearables market long after Fitbit. But in each case, Apple waited, watched, and built a product that actually made sense.
Right now, there’s no real market for AI, only investment. OpenAI, Microsoft, Google, Amazon, and Anthropic are spending billions on training and inference infrastructure.
They’ve released tools that can generate text, images, audio, and code. But these tools haven’t become profitable products.
OpenAI’s ChatGPT saw rapid growth in 2023 and recent data indicates over 400 million weekly users. That’s mostly via Microsoft integrations or freemium models.
High interest doesn’t equate to high revenue. Hell, OpenAI gladly took the opportunity to plug into Apple Intelligence for free. Sometimes exposure does pay rent, at least if you shuffle numbers around.
AI is unlikely to disappear, but Apple hasn’t failed. It’s not late to the game because there is no game.
Some people have just believed the nonsense propagated by stock analysts, and AI company CEOs to hype up shareholders and investors.
