Ternus Puts Apple’s AI Strategy in Product Hands

Apple’s hardware teams now have a product problem that cannot be solved by making the next device thinner, faster, or easier to hold. With John Ternus taking over as chief executive on September 1, AI and a rebuilt Siri become the company’s top priority, according to Bloomberg’s report. The immediate consequence is organisational before it is technical: product managers who once could treat intelligence as a feature layered onto a device will increasingly own how that intelligence behaves across the whole product system.

That is a serious change in the job. Ternus arrives with a hardware engineering background and a set of upcoming device launches in front of him, but the mandate attached to the role is not simply to protect Apple’s hardware strength. It is to make the assistant useful again, and to connect it to the places where people now expect generative AI, meaning software that can produce or transform content in response to an open-ended request. Nobody outside Apple’s labs knows how far the rebuild has gone, or whether the company has closed the gap with its competitors. We do know where the pressure now sits.

For hardware-centric PMs, the old boundary between the device roadmap and the assistant roadmap is becoming difficult to defend. A new form factor can change what a person asks, when they ask it, and what a satisfactory answer looks like. A foldable iPhone, which Bloomberg reports may be among the hardware shown at Apple’s expected September 9 event, is not merely another screen configuration if Siri is meant to understand the user across changing contexts. The hinge is a hardware decision. The assistant’s memory of the task, its handoff between screens, and its confidence about what to do next are product decisions that belong in the same room.

The assistant becomes the surface

The most important shift is not that Siri gets more capable. It is that Siri becomes a measure of whether Apple’s devices work as one product. An assistant that can respond to a request but cannot carry the user’s intent from one device state to another is still a collection of features. An assistant that knows when to answer, when to ask, and when to act changes the role of the product team from shipping interfaces to designing a continuing interaction.

Consider a simple request: a person begins planning something on one screen, folds the device, and expects to continue without explaining the task again. The visible interface may change, but the product promise does not. The PM has to decide what context persists, which action is safe to take without confirmation, and how the user recovers when Siri misunderstands. Those questions sit partly in conversation design, partly in platform behaviour, and partly in hardware integration. They cannot be handed off to an AI team at the end of the roadmap without creating a fragmented experience.

That is why “new Siri” matters as a product mandate rather than a feature announcement. The assistant is a front door into the operating system, and every weak handoff makes the underlying product feel less coherent. If Siri gives a polished answer but cannot complete the next step, the failure is not confined to language quality. If it acts too eagerly, the failure is trust. If it works on one device but not another, the failure is the architecture of the product as a whole, even if each individual component performs acceptably.

A PM working on a device therefore needs to own more than the assistant’s placement or invocation method. They need a view of the end-to-end journey: the request, the interpretation, the response, the action, the confirmation, and the recovery. That does not mean every hardware PM must become a model researcher. It does mean they must be able to set the product contract with the teams building the model, the operating system, and the physical interaction. “Siri is available here” is no longer a sufficient requirement.

Hardware now carries intent

New form factors make this unavoidable because they alter the conditions under which people use an assistant. A larger or differently shaped device can create room for richer results, but it can also introduce new ambiguity. Is the user asking for a summary to read, an action to take, or a draft to edit? Should the response occupy the whole display, remain alongside another task, or wait for confirmation? These are not decorative choices. They determine whether the assistant feels like a useful collaborator or an interruption.

The September product event is therefore a meaningful test, even before Apple has announced what it will show. Bloomberg reports that a redesigned Siri is expected alongside new hardware, including a possible foldable iPhone. We should be careful here: an expected launch is not evidence of a finished strategy, and a redesigned interface is not proof of a capable assistant. But the timing makes the product question plain. Apple is being asked to demonstrate that its AI ambition belongs inside the device experience, not beside it as a separate destination.

The counterpoint is obvious, and it matters. Ternus’s engineering background could be an advantage precisely because AI has to become dependable in a physical product. He may be better positioned than a purely software-minded leader to insist that latency, battery behaviour, privacy expectations, and interaction details are part of the promise. Yet hardware discipline alone will not produce a good assistant. A device can be beautifully integrated and still fail if Siri does not understand the user’s intent, cannot complete a task, or offers no graceful way back from an error.

This is where competitive catch-up becomes a product management problem rather than a press-release problem. Apple does not need to prove that it can speak about generative AI. It needs to show that the capability is useful often enough, reliable enough, and connected enough to earn a place in daily habits. The relevant comparison is not a model benchmark. It is the moment when a user chooses whether to ask Siri or reach for another product because they already know which one will finish the job.

Roadmaps need a new owner

The practical response for PM leaders is to stop treating AI as a dependency that arrives from a central group with a launch date. If Siri is now a top company priority, every hardware roadmap that touches it needs explicit ownership for the assistant experience, from discovery through post-launch learning. That includes the use cases the form factor creates, the tasks it should refuse, the points where a person must remain in control, and the evidence that would justify expanding the assistant’s authority.

The measures should follow the product promise. A team can track whether people invoke Siri, but invocation alone says little. The harder questions are whether users complete the intended task, whether they repeat the behaviour, whether they correct the assistant, and whether they abandon it after a failure. For a new device, those signals should be read alongside physical context: when the user is holding the device one-handed, switching screens, moving between visual and spoken interaction, or trying to resume an interrupted task. The point is not to create a larger dashboard. It is to connect model behaviour to the moments the hardware was designed to improve.

There is also a quieter ownership change. Hardware PMs will need to negotiate what the assistant is allowed to do on behalf of the user, not only what it can display. That means defining confirmation patterns, visible state, fallback behaviour, and the boundary between an answer and an action. A useful assistant may save a person from navigating a menu, but the product still has to make its decision legible. If the user cannot tell what Siri understood or what it changed, the convenience will not survive the first costly mistake.

The temptation will be to make the fall launch carry too much meaning. Apple may show a compelling demo, and the market may read that as evidence that the company has solved the wider product problem. That would be premature. The real work begins after the scripted path, when teams see the requests they did not anticipate, the device states they did not model, and the moments when users decide that speaking to a machine takes more effort than doing the task themselves.

Ternus’s appointment makes that work harder to hide. AI is no longer a side bet that can be evaluated separately from the hardware business, and Siri is no longer only an assistant team’s backlog. The device, the model, the interface, and the user’s trust now arrive as one product. If you lead a hardware roadmap, ask the uncomfortable question before the next concept review: when Siri fails on your device, who owns the failure? If the answer is still “the AI team,” your roadmap has not caught up with Apple’s new priority.