Apple pushes for an agentic Siri: the iPhone, more than the model, may be its edge in a race it trails

Apple is speeding up its Siri roadmap. Bloomberg's Mark Gurman reports that CEO John Ternus is pressing Craig Federighi, the executive who leads the company's AI efforts, to deliver a capable AI agent. An agent here means an assistant that carries out tasks rather than only answering questions. The benchmarks Apple has in mind are OpenAI's Dots and Meta's Muse. Gurman says Federighi has already shipped a working version of Siri AI on a very short timeline. But he also reports that the company has no illusions about keeping pace with its rivals.

Gurman expects the work to come in stages. The first, planned for iOS 27.3, gives Siri fine-grained control over Apple's own apps. Examples include rotating and cropping images in Photos and organizing tabs in Safari. The bigger step is integration with third-party apps, which Gurman expects in the spring. That would let Siri place an order on Grubhub or book a table on OpenTable. After that comes the real goal: tasks that continue over time, not single commands. Gurman describes how Dots, Muse and Grok already work. They keep track of goals, work out the next step and keep going without the user starting each step. He believes Siri could start working that way as early as next year, "not that it has much of a choice, given current consumer and market demands."

That last phrase stands out. A few years ago, Apple could afford to arrive late and arrive polished. That was its strategy with smartwatches and contactless payments. With agents, the timing has changed. Users now compare Siri every day with assistants that already handle multi-step tasks. The gap is no longer a missing feature. It is a different category of product. By Gurman's account, Apple knows it is behind and is acting on that.

Our thesis: Apple's best asset in this race is not the most capable model. It is the platform. In earlier analyses we argued that the AI war is moving toward distribution and integration, and that whoever controls the "plumbing" wins. Few companies control more plumbing than Apple. It runs the operating system, the App Store and the permissions model on hundreds of millions of devices. An agent that can open Photos, Safari, Grubhub and OpenTable from inside the system has access that an outside assistant has to negotiate for. The hard part is the model and the reasoning across many steps, and on that front even Apple's own leadership seems to accept that it trails.

A second point matters just as much. When an assistant becomes an agent, the risk changes from a wrong answer to a wrong action. A Siri that crops a photo badly is a minor annoyance. A Siri that places an order or books a table for you is spending money and making commitments on your behalf. If it takes on ongoing tasks, as Gurman expects, the room for error grows. That is why the staged rollout makes sense: first Apple's own apps, then third-party apps, then autonomy. Apple has built its brand on privacy and control. Its real test will be explaining clearly what the agent can do on its own, when it asks for confirmation, and how a user can undo an action.

Our reading: in the short term, this is a company catching up under pressure, with schedules that come from leaks rather than official announcements. They should be treated as plans, not commitments. In the long term, the direction is good for users. Agents that take over the small tasks of digital life, such as orders, bookings and organizing, give people back time and attention. If Apple can combine that convenience with strong control and privacy safeguards, the platform that arrived late could still set the standard for how an agent should behave on a personal device. Getting there will require more than speed.