Inc.
After years of AI embarrassment, Apple may have finally made Siri worth using again.
The Most Important Thing Apple Got Right With Siri AI
Author: Jason Aten
For two years, Apple has had a very real problem. The story of the company and its AI efforts has been basically that Apple was hopelessly behind. Ever since it rolled out Apple Intelligence at WWDC in 2024, the narrative has been that the company made big promises, but delivered on almost none of them.
Going into WWDC this month, the expectations were high. If Apple was ever going to change the impression that it had no coherent strategy for making its Siri voice assistant good, it had to be now.
At this year’s WWDC, Apple did exactly that with Siri AI. More than a rebrand, it’s a full rebuild, running on models built with help from Google’s Gemini. On the one hand, it’s an admission that couldn’t get there on its own—at least, not on a timeline that anyone would tolerate. On the other hand, the fact that Apple was willing to partner with anyone for what is arguably its most important piece of primary technology is a sign that it is taking this seriously.
The good news is, after a few weeks of using the beta version of iOS 27, it seems like Apple has finally done what has so far been impossible: it made Siri good. There are still some issues, but that’s because it’s a beta. This isn’t meant to be a Siri AI review, as much as a look at how Apple managed to do what it promised.
It turns out that if an LLM has access to all the information on your device, it can be pretty useful. Siri AI is useful, and it mostly just works. There is finally a standalone Siri app, the way you’d use ChatGPT or Gemini, and it saves past conversations and syncs across iPhone, iPad, and Mac through iCloud.
It’s useful because it can read your email and text messages, and can even see what’s on your screen. Get an email with flight details, hold the side button, and Siri can add the event to your calendar and message someone the arrival time without you copying a single field. I asked it about a conversation with my daughter, and without giving it any context, it told me about a conversation we had had a few days earlier.<figure class=”wp-block-image size-large”>
</figure>
If those capabilities sound familiar, they should. Apple promised most of them in 2024.
Back then, the personalized Siri — the one that understood your context and acted across your apps — was the headline of the entire keynote. Then it didn’t ship. Apple delayed it, pulled an ad it had already run, and in May agreed to a reported $250 million settlement with customers who said they’d bought iPhones for features that never arrived.
The bar this year was to deliver on whatever promises it made. Apple cleared that bar in the most un-flashy way available. Look at how the keynote was built. Apple led with fixes before features — faster app launches, quicker AirDrop, iOS 27 running all the way back to the iPhone 11 — and folded the better Siri into a long list of improvements rather than staging it as the second coming. Two years after over-promising on this exact product, Apple showed off only what you could try out for yourself. That was a choice, and it was the right one.
Every keynote now opens with a company insisting its AI will change your life, usually while demoing something that works in one controlled clip and nowhere else. The gap between the promise and the product is the hallmark of the AI era. That’s because most AI companies are startups trying to impress investors in order to raise money. Hype is the only thing that matters.
That’s not the case for Apple. Its IPO was almost 46 years ago. Apple has to ship products that people love, and almost no one loved old Siri.
Craig Federighi spent part of the keynote drawing the contrast on purpose. He said some companies are racing forward to build AI for the sake of AI, and that privacy in AI is non-negotiable. You can read that as marketing. You can also read it as Apple announcing that it has stopped trying to win the hype cycle and started trying to make something people actually want to use.
Honestly, the features are good. Whatever concoction of Apple’s Foundation Models trained by Gemini under the hood seems to work. The standalone app is table stakes at this point. The on-screen awareness is the thing people actually wanted back in 2024.
The features, however, are not why Siri is a success. No, the real thing Apple got right this year is that it promised exactly what it can deliver and dared everyone to hold it to that.
Credits: TCA, LLC.