Apple's AI reset is no longer about building the smartest model in-house. It is about making Siri useful enough that users stop caring which model is underneath.
Apple heads into WWDC 2026 with a problem that is unusually simple for a company this large: people remember what it promised, and they have not yet seen enough of it work. The company's June 8 keynote, which opens its June 8-12 developer conference, is expected to put a rebuilt Siri and a broader Apple Intelligence strategy back at the center of the story.
That matters because this is not just another software refresh. According to Bloomberg's latest reporting, Apple's sharper AI pivot followed a private internal meeting led by software chief Craig Federighi, who took direct oversight of the company's AI organization after the first Apple Intelligence rollout failed to land with the force Apple expected. Federighi reportedly pushed for closer collaboration between software and AI teams while making clear that the pace had not been good enough.
The result is a strategy that looks much less like old Apple. Instead of insisting that its own foundation models carry the whole experience, Apple is preparing Siri for a world where large language models are treated more like infrastructure. Gemini, Claude, ChatGPT and Apple's own systems can become interchangeable pieces, routed into the user experience when they are the best tool for the job.
For years, Apple won by owning the full stack. Hardware, software, services and distribution all worked better because they were designed together. AI has challenged that formula because the most visible consumer breakthroughs came from companies built around model scale, cloud infrastructure and fast product iteration. OpenAI, Google and Anthropic moved in public. Apple moved carefully, then delayed the part users cared about most.
The reported Siri plan is Apple's answer to that mismatch. MacRumors, summarizing recent Bloomberg details ahead of WWDC, said iOS 27 is expected to let rival chatbots integrate with Siri through Extensions, expanding beyond the existing ChatGPT handoff and allowing users to send requests to services such as Claude or Gemini. Users may also be able to choose third-party AI tools as defaults for features like Writing Tools and Image Playground.
That is a big concession, but it is also a practical one. If the model layer changes every few months, Apple does not need to win every benchmark to win the user relationship. It needs to control identity, permissions, context, privacy boundaries and distribution. In that version of the market, OpenAI and Anthropic still matter, but they risk becoming premium engines inside someone else's interface.
This is why the move matters beyond the iPhone. OpenAI and Anthropic have spent heavily to position their models as the trusted layer for companies that need reliability, safety and strong reasoning. If Apple normalizes a chatbot-agnostic architecture on hundreds of millions of devices, buyers may become more comfortable swapping models based on cost, policy, task or geography. The model remains valuable. The lock-in weakens.
Federighi's reset has to show up in the product
Apple's organizational change is only useful if it produces software that feels obvious once people use it. The expected Siri upgrade is reported to include more natural conversation, deeper personal context, stronger app integration and a standalone Siri app that looks closer to modern chatbot products. That would mark a sharp turn from Apple's earlier reluctance to make Siri feel like ChatGPT.
The Google partnership is the clearest sign of how far Apple has moved. Apple and Google confirmed in January that Gemini models would support the next generation of Apple Foundation Models, with Tim Cook later describing the arrangement as a collaboration while saying Apple Intelligence would continue to run through Apple devices and Private Cloud Compute. MacRumors has reported that the deal could cost Apple about $1 billion a year, though Apple has not disclosed the terms.
That cost is small compared with the price of losing the assistant layer. Siri sits at the entrance to messages, calendars, apps, files, photos, search and device actions. If Apple can make it genuinely useful, it gives developers a new surface to build around and gives users a reason to see Apple Intelligence as part of the operating system rather than a feature folder.
The harder question is whether this closes the gap with Google and Microsoft. Google has spent the past two years pushing Gemini across Search, Android, Workspace and developer tools. Microsoft moved quickly with Copilot across Windows, Office and Azure. Apple still has the advantage of device integration, but it no longer has the luxury of proving AI through polished demos alone.
WWDC will therefore be judged less by how futuristic Siri sounds and more by how specific Apple gets. Developers need APIs, timelines and clear rules for what third-party models can do. Investors need evidence that Apple Intelligence can support device demand and services growth. Users need to see Siri complete real tasks across apps without turning every request into a handoff or an apology.
Apple's best path is not to pretend the delay did not happen. It is to make the delay look like the moment it stopped treating AI as a feature race and started treating it as a platform decision. If Federighi's reset works, Apple will not have to own the best model every week. It will have to own the place where those models become useful.
Also read: AI stock deals are testing how much Wall Street can absorb • Trump’s AI memo redraws the defense market • Britain is becoming the first buyer for homegrown AI chips
