Apple has spent two years being described as an AI laggard, but a job listing posted on its careers site on November 17 suggests the company is building the same kind of technology its rivals already ship. The posting, titled Senior Engineering Manager, Agentic AI — Proactive and located in Cupertino, described work on a "general-purpose computer-use agent" — software that operates a Mac the way a person does.
The listing's own framing was unusually direct for Apple. It asked candidates to imagine an AI agent that can use a computer the way a capable person does: browsing the web, working across apps, running terminal commands, writing and debugging code, and calling outside services, choosing the right tool at each step. The role involves coordinating with the Siri, UI Frameworks and Apple Foundation Models teams, and the agent is expected to plan and execute multi-step tasks and to coordinate specialized subagents. Safeguards — permission models, sandboxing and transparency — were named explicitly as part of the design brief.
Bloomberg's Mark Gurman reported on November 19 that the project is in early stages and that Apple has eyed introducing the service as soon as the end of 2027. Shortly after the reporting, the listing disappeared from the careers site — a common Apple practice to avoid telegraphing unannounced products — and the company did not respond to press inquiries. Gurman has previously reported that CEO John Ternus, who took over from Tim Cook on September 1, has been pressing software chief Craig Federighi to move faster on agentic AI.
The competitive context explains the urgency. Meta's Muse agent has become one of the most-downloaded apps on Apple's own App Store, OpenAI's Dots, Anthropic's computer-use tooling and Google's Gemini computer use are all shipping, and rivals' assistants now occupy most of the top free-app slots. Apple's current Siri AI, released with iOS 27 in September, can read the screen and perform limited in-app actions, but it is not an agent that independently plans and completes long tasks across applications.
Notably, Apple is simultaneously making it harder for everyone else's agents to control a Mac. macOS 27.3 beta 2, released November 16, separates system-control and accessibility permissions, adds an on-screen indicator when one app controls another, and requires more narrowly scoped grants. Digital Trends, citing Bloomberg, counts the steps to enable ChatGPT's computer-use tools in the new beta at 14, up from 4 — with warning screens asking users to reconsider before granting full access to an outside agent. Apple frames the changes as security; some AI companies and users see an attempt to slow competition.
Nothing here is a product announcement: there is no name, no feature list and no confirmed date, and Apple has said nothing publicly. What the listing does establish is intent and structure — a named team, a defined capability set, and an internal timeline. If the reported late-2027 window holds, Apple would be entering the computer-use agent market roughly three years after its rivals, but with something none of them have: default control of the operating system those agents need to run on.
Comments (0)
Log in to join the discussion
Log InNo comments yet