The startup is called Poke, and it has done something no other company has managed: get an AI agent approved to run on Apple’s Messages for Business platform. The approval, reported by TechCrunch, makes Poke the first third-party agent to operate inside Apple’s business-messaging service, where it can schedule meetings, manage calendars, track fitness, control smart-home devices and edit photos for customers who reach a business through iMessage.
Poke launched in March and says it has processed more than 100 million messages since. It charges businesses for the automation it performs, and it pays Apple a per-user fee to operate on the platform — a revenue arrangement that gives Apple a cut of the agent economy without building the agents itself. The company recently raised an additional $10 million at a $300 million valuation.
The approval is modest in scope and significant in structure. Apple has long resisted letting third-party software reach deeply into the iPhone experience, and its business-messaging platform has been a tool for customer service rather than a marketplace for software agents. Letting Poke run there — and taking a fee for it — is the company’s first concrete answer to a question every platform owner is facing: when AI agents start doing business on behalf of customers, who collects the toll?
The other answer arrived the same week, and it involves hardware. Analyst Ming-Chi Kuo reported that incoming Chief Executive John Ternus has shelved the successors to the Vision Pro, canceling the planned second-generation headset and the lighter Vision Air. In their place, Apple’s roadmap now contains two smart-glasses products: an AI-powered pair without displays, targeting 2027, designed to compete with Meta’s Ray-Ban glasses; and a more advanced augmented-reality pair with optical displays, slipped to 2029. The company’s wearable roadmap, which at one point covered seven headset-related products, has been cut to two.
The two moves define the shape of the Ternus era, and they point in opposite directions. In software, Apple is being careful: one agent, one platform, a fee structure that keeps control in Cupertino. In hardware, the company is being decisive: the headset that its previous leadership treated as the next iPhone has been quietly retired, and the resources behind it are moving to a product category where the company is behind — Meta shipped its AI glasses years ago and has the distribution to make them ordinary.
Vision Pro’s retreat was gradual but unmistakable. The $3,499 headset impressed reviewers and sold in disappointing numbers, and the follow-up products that would have justified the platform’s existence have now been canceled before reaching production. The pivot is a bet that the future of wearable computing looks like glasses, not goggles — a bet Meta made early, a bet Apple is now making with full commitment.
Ternus, who has run hardware engineering and is set to take over as chief executive, has been described by colleagues as more focused on product decisions than on grand narratives. The two announcements fit that description: no vision statements, just a platform opened a crack and a product line redirected. Whether the bet works depends on execution in both directions — whether Poke’s approval becomes the beginning of an agent marketplace or a one-off, and whether Apple’s glasses, still two years away, arrive with the polish that Vision Pro lacked.
For the developers who watched Apple’s history with third-party AI, the Messages for Business decision is the more telling signal. Apple has been slower than its rivals to embrace agentic AI, in part because its privacy architecture makes deep integration hard and in part because the company’s instinct is to control. The Poke arrangement threads that needle: the agent runs on Apple’s platform, pays Apple a fee, and works within the rules Apple sets. It is an opening, not an open door.
For the hardware world, the glasses pivot is the bigger story. Apple’s brand has rested on entering categories late and winning them, and the company is now doing exactly that in wearables — against a competitor that has a multi-year head start and a product that already sells in the millions. The stakes are unusually high for a company that rarely bets on a single new category: the glasses program is now the clearest expression of where Apple believes personal computing is heading after the phone.
The two lines will develop on different clocks. Poke’s approval takes effect as businesses decide whether to put AI agents in front of their customers. The glasses arrive in 2027 at the earliest, and their success depends on hardware, software and an ecosystem that does not yet exist. What is already settled is the direction: Apple is opening its software platform to AI agents one careful step at a time, while closing its hardware past and moving everything behind a new one. The Ternus era will be defined by whether both bets pay off in the same window.
