Apple accelerates Siri development: agentic features in iOS 27.3, third-party apps in spring

For years, Siri could set timers and answer questions about the weather, while competitors were teaching their assistants to get things done. Now Apple is scrambling, according to Bloomberg's Mark Gurman. CEO John Ternus is pushing Craig Federighi: the company needs a powerful AI agent capable of competing with OpenAI's Dots and Meta's Muse.

Apple itself appears to have no illusions, judging by Gurman's account. Federighi put together a working version of Siri AI on a very tight schedule, but the company is not keeping pace with its rivals.

What to expect, and when

Gurman identifies two major improvements coming in the next few months. The first will arrive with iOS 27.3: Siri will gain fine-grained control over Apple's native apps. This means taking actions inside an app, rather than simply responding to an "open the app" command. Users will be able to rotate and crop an image in Photos or organize tabs in Safari.

The second, more significant improvement is expected in spring: integration with third-party apps. Gurman cites examples such as placing an order through Grubhub or booking a table through OpenTable. For the ecosystem, this matters more than cropping photos, because it is where the assistant starts carrying out real tasks on the user's behalf.

How an agent differs from an assistant

Apple's bigger bet lies further ahead: teaching Siri to handle ongoing tasks, rather than just one-off commands. As Gurman describes it, Dots, Muse and Grok can keep a goal in memory, determine what to do next and continue working toward the result without instructions at every step.

A simple illustration follows, using our own example rather than one from the source. Asked to "find dinner for Friday," an assistant will show a list of restaurants and leave the rest to you. An agent will take on the whole goal: check availability, make a reservation, remind you about the visit and suggest an alternative if the table is no longer available. You specify the outcome, rather than the steps.

Gurman believes Apple could make Siri more like an agent as early as next year. In his view, the company simply has no choice, given consumer and market demands.

What this means in practice

For iOS developers, the main question is exactly how Siri will access functions in third-party apps. The source does not explain this. It is reasonable to expect Apple to require developers to explicitly describe which actions their apps expose to the system, so it is worth preparing those workflows in advance. This is our assumption, not an established fact.

For product managers, the implication is starker. If your product handles orders, reservations or appointments, Siri could become a new entry point where users never see your interface at all. Competition would then center on the quality of the outcome and the availability of actions, rather than screen design. For founders, this offers both a chance to reach a channel spanning hundreds of millions of devices and a risk of becoming an anonymous backend behind the assistant's voice.

Limitations

Everything above rests on Gurman's sources, rather than an announcement from Apple. He himself frames the timing cautiously, particularly for agentic capabilities: "could" and "as early as next year." It remains unclear how Apple will handle permissions, payments and errors when an agent acts on someone's behalf—precisely the issues that cause agentic products to stumble.

The takeaway: Apple acknowledges it is behind and is choosing a phased approach. First its own apps, then third-party apps, then autonomous tasks. For the market, this signals that an agentic interface is becoming a standard expectation, even for the most cautious platform.