Search
Tech

Apple job listing reveals plans for its own AI agent that can run a Mac

The posting describes software that browses, codes and works across apps. Bloomberg says a launch could come as soon as late 2027, while Apple tightens rules for rival agents.

Aerial view of Apple Park, Apple's ring-shaped headquarters campus in Cupertino, California.
File photo. Photo: Daniel L. Lu (user:dllu) / Wikimedia Commons, CC BY-SA 4.0

Key Takeaways

  • An Apple job listing, since removed, describes an agent that browses the web, runs commands, writes code and works across apps.
  • Bloomberg reports Apple has eyed a launch as soon as the end of 2027, but the project is still early.
  • macOS 27.3 will add warnings that discourage users from giving third-party agents full access to their Macs.

Apple is building its own AI agent that can operate a computer on a user's behalf, according to a job listing first spotted by Bloomberg. The posting describes software that browses the web, moves between apps, runs commands, writes and debugs code, and calls online services, picking the right tool at each step.

The listing, for a role titled "Senior Engineering Manager, Agentic AI – Proactive," has since been taken down. Its opening line sets the goal plainly: "Imagine an AI agent that can use a computer the way a capable person does." A few sentences later comes the pitch only Apple would write: "It does all this privately and with the polish people expect from Apple."

Bloomberg reported that Apple "has eyed introducing the service as soon as the end of next year," which would mean late 2027. The same report cautioned that work on the in-house agent "remains in the early stages." TechCrunch said Apple did not respond to its request for comment.

What the listing says Apple is building

The posting goes beyond a smarter Siri. It describes a team of engineers, applied AI researchers and systems specialists already working on what it calls "general-purpose computer-use agents." Those agents would plan and complete multistep jobs across apps, browsers, command lines, code and APIs, and could hand pieces of a task to specialized sub-agents along the way.

The manager Apple hires would work across three groups: Siri, UI Frameworks and Apple Foundation Models, the team behind the company's in-house AI models. The listing also says safety, privacy and security must be "central to the design."

That combination is telling. Pulling in the UI Frameworks team suggests Apple wants an agent that understands the building blocks of Mac and iPhone apps from the inside, rather than one that only looks at screenshots and clicks where it guesses a button might be. If Apple can pull that off, its agent could be more reliable inside Apple's own software than outside tools are. That advantage would not extend to the web, where Apple has no special access.

The listing fits with other recent reporting. Bloomberg said days earlier that Apple wants to make Siri act more like an agent as soon as 2027, carrying out tasks proactively instead of waiting for a user to start each step. According to Bloomberg, possible new Siri skills include rotating and cropping images in Photos and organizing Safari tabs.

Apple started rolling out its AI-upgraded Siri in September. The assistant can answer questions about what is on the screen and take actions across apps. With the upcoming iOS 27.3 update, Apple is also offering developers a way to bring Siri AI into "all apps" through a Developer Preview, which should eventually let Siri work inside far more third-party software. What Siri cannot yet do is act on its own initiative, the defining trait of the newer proactive agents.

Rivals already have a head start

Apple is arriving at a crowded table. Meta released its Muse agent in September, and Sensor Tower data puts it among the fastest-growing apps in U.S. history, with more than 10 million downloads in 40 days. On Nov. 19 it ranked No. 2 among free apps in the App Store, trailing only TikTok, according to 9to5Mac. OpenAI followed with Dots, Anthropic launched Claude Duet, and Google rebranded its Gemini Spark agent as Tasks and added features.

On the Mac specifically, Muse, ChatGPT and Perplexity can already take control of the computer through their desktop apps, and Google has just begun letting Gemini do the same. In other words, Mac owners who want an agent driving their machine have several options today, and Apple's own is at least a year away by Bloomberg's most optimistic timeline.

That gap is the central problem. Apple shipped its long-promised Siri overhaul only after the industry had already moved on to autonomous agents, and a late-2027 target leaves plenty of room for the same thing to happen again. The listing's emphasis on privacy and polish reads like Apple's familiar answer to being late: arrive second, but arrive better. That strategy has worked for hardware. It is a harder sell in AI, where rivals' tools improve month to month and users build habits quickly.

Leadership pressure may help. John Ternus, Apple's new CEO, has reportedly been pushing software chief Craig Federighi to move faster on AI. A public hiring push for a senior agent lead is consistent with that, though a job posting is a long way from a shipping product.

Tighter rules for third-party agents

While Apple builds its own agent, it is making it harder for outside ones to take full control of a Mac. Bloomberg reported that macOS 27.3 will show warnings discouraging users from granting complete access to third-party agents.

The groundwork was laid in October, when Apple told developers it would add controls so that people who truly want to give an app what it called "this extraordinary level of access" can do so only through "very explicit user action." The macOS 27.3 beta tightens the Full Disk Access setting and gives users more granular control over which apps and data third-party agents can reach.

There is a real security case for this. An agent with full access to a computer can read files, send messages and run commands, and a mistake or a compromised tool could do serious damage. Making users think twice before handing over that power is defensible on its own terms.

The timing is the harder part. Whatever the reasoning, the practical effect is that rival agents on the Mac will face more friction just as Apple prepares a competitor that, as a built-in system feature, is unlikely to carry the same warnings. The question worth watching is whether Apple's own agent will be held to the same consent standards it is now imposing on tools like Muse and ChatGPT.

For now, the clearest near-term change is the one Mac users will see first: more prompts and tougher permission screens for third-party agents in macOS 27.3. Apple's own computer-use agent, by Bloomberg's account, is still at the stage of hiring the person who will lead it.

Sources

Ethan Kim

Ethan Kim writes about consumer technology, including phones, computers, and connected devices. His coverage examines product announcements, specifications, and the tradeoffs behind new hardware. He focuses on turning technical details into useful context, helping readers understand how devices differ and which changes represent a meaningful step from earlier generations.

View all stories