Apple is putting an AI agent at the center of 2.5 billion devices
On May 28, 2026, Bloomberg reported that the centerpiece of Apple’s iOS 27 is a rebuilt Siri. The reporting describes a chatbot-style assistant that keeps a conversation history, accepts text, voice, photos, and documents, and performs actions across the phone: composing messages, adding calendar entries, opening apps, searching notes. A new “Search or Ask” panel slides down from the top of the screen and returns formatted answer cards. The Camera app gains a dedicated Siri mode. For some questions, Siri can hand off to outside models, including Google Gemini, Anthropic Claude and OpenAI ChatGPT.
Apple will show this at its developer conference on June 8. It will ship to a base of 2.5 billion active devices. The interaction model is moving from “open an app and tap” toward “ask an assistant and get an answer.”
Here is a signal: the AI assistant is becoming the layer between a person and the information they want, and the answer it gives is assembled by a model from data the person never sees.
What actually changes for the user
The clearest way to read this is from the perspective of the person holding the phone. A few things change in how the device is used day to day.
The starting point moves. Instead of unlocking the phone and looking for the right app, you swipe down for a “Search or Ask” box, or talk to a Siri that now lives in the Dynamic Island. The phone returns an answer card rather than a list of apps to open.
Siri keeps the thread. The new chatbot-style app stores a conversation history, so you can ask a follow-up without repeating the context, the way you already do in ChatGPT.
You can show it things. You can hand Siri a photo or a document and ask about what is in it. In the Camera, a dedicated Siri mode lets you point at an object and ask what it is.
One sentence can trigger several steps. Shortcuts can be built by describing them in plain language. Bloomberg’s example is a routine that starts a music playlist and texts your partner an ETA when you begin driving home.
Editing happens by description. In Photos, “Reframe” and “Extend” change a shot’s framing or perspective, and you can ask for edits such as cropping or colour changes by voice or text.
The answer often comes from Google. For the harder questions, the reasoning is reportedly handled by Google’s Gemini, running inside Apple’s private cloud. The label says “Siri.” Much of the thinking is Google’s model.
The wider effect is a habit. Once tens of millions of people get used to asking a question and receiving one direct answer, the same expectation carries over to everything they look up, including which product to buy and which company to trust.
A fair question: can Apple’s new Siri keep up?
It is worth holding the renders at arm’s length because Apple has shown an AI-driven Siri before and not shipped it.
At WWDC in June 2024, Apple previewed a personalised Siri that could use your personal context and take actions across your apps. The feature was promoted in iPhone 16 ads. Then it slipped. In March 2025, Apple said the work would take “longer than we thought” and pushed it into “the coming year.” Apple’s own Siri chief described the delays as “ugly and embarrassing”. Apple later agreed to pay $250 million to settle a class action that accused it of false advertising over features it marketed but did not deliver.
The bigger tell sits under the hood. In January 2026, Apple and Google announced that Gemini would power the rebuilt Siri and Apple’s foundation models, with Apple reportedly paying around $1 billion a year for a custom version. Apple’s statement called Google’s technology “the most capable foundation” for its models. Several commentators read that as an admission that Apple’s in-house models had fallen behind.
So iOS 27 arrives with a strong feature list on paper and a competitor’s model doing the heavy lifting. The renders look polished. Apple’s previews have looked polished before, and the gap between the keynote and the shipped product has been the problem.
The open question is whether the version people actually use matches what they now expect from ChatGPT and Gemini, or lands as another late, underwhelming release, the kind of AI feature that demos well and then becomes more of the “slop” people learn to scroll past. We will know more after June 8, and more still once it ships. Betas are expected over the summer and the release in the fall, and the most personal features have a habit of slipping.
Why this matters now
The move toward asking an assistant first is already underway, with or without Apple. ChatGPT passed 900 million weekly users in February 2026 and counts more than 9 million paying business users. Apple is about to put an agentic assistant, its own or Google’s, depending on how you read it, in front of 2.5 billion devices.
For any company whose customers ask questions before they buy, the relevant change is the same regardless of which assistant wins. More of those questions get answered by a model, and the answer is assembled from public data the company does not control. That happens whether the model is Gemini inside Siri, ChatGPT, or Perplexity. The assistant layer is worth watching closely now, while the habit is still forming, rather than after it becomes the default.
The interface on the most personal device is turning into an assistant you talk to. iOS 27 is Apple’s attempt to catch up, built on Google’s model and a long record of missed dates. It may finally land. It may slip again. Either way, people are learning to ask first and read links later, and that habit is the part worth paying attention to.
