Apple posted a job listing on November 17 for a Senior Engineering Manager, Agentic AI – Proactive. The description outlines an autonomous agent that can use a computer the way a capable person does: browse the web, work across apps, run commands, write and debug code, call services, and choose the right tool for each step. The listing says this should happen privately and with the polish people expect from Apple. Bloomberg first reported the hiring push, and multiple outlets have since confirmed the details from the same posting. [2] [6]
What the agent is designed to do
The job summary describes a system that plans and carries out multi-step tasks across apps, browsers, command lines, code, and APIs. It can coordinate specialized sub-agents along the way. Apple calls these “general-purpose computer-use agents” in the listing. That phrasing matters: it signals an architecture closer to OpenAI’s Dots or Anthropic’s Duet than to today’s Siri, which depends on developers exposing specific actions through App Intents. The new agent would navigate user interfaces directly and decide how much autonomy to use for a given task.
How it differs from Siri and third-party agents
Current Siri integrations require apps to opt in via App Intents. The agent in the job listing would operate at the UI level, moving between browsers, apps, terminals, and services without needing per-app hooks. That puts it in the same category as the computer-use agents from OpenAI, Anthropic, Meta, and Google. The difference Apple emphasizes is privacy and polish. The listing explicitly calls for clear permissions, sandboxing, and visibility into what the agent is doing. Those requirements line up with the granular permission changes Apple is already testing in macOS 27.3 beta, where users can limit which protected folders an app can access and require automation tools to ask before controlling specific apps.
The privacy and permission architecture
Apple’s approach appears to treat the agent as a first-party automation layer that respects the same sandbox boundaries as any other app. The macOS 27.3 beta increases the authorization steps for third-party AI agents from four to fourteen, a move Mark Gurman described as Apple citing security while some users see anti-competitiveness. For Apple’s own agent, the job listing demands “clear permissions, sandboxing, and visibility” — suggesting the company wants to set a higher baseline for transparency before it ships. The engineering lead would work across the Siri, UI frameworks, and Apple Foundation Models teams, indicating deep system integration rather than a bolt-on service.
Timeline and competitive pressure
The listing gives no launch date or product name. Bloomberg’s Mark Gurman has reported Apple could introduce agent-like Siri abilities as soon as late 2027, while other coverage cites guesses pointing to 2027. Development language in the posting suggests early stages. Meanwhile, Meta’s Muse reached 10 million downloads in 40 days and has held the No. 2 spot among free App Store apps. OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Tasks are also in the top five. Apple’s hiring push follows the breakout success of these agents and the proliferation of tools that let AI control Macs and write code. Shares of Meta and Anthropic slid after Bloomberg reported Apple’s efforts.
What this means for users and developers
For users, the promise is a single agent that can handle a workflow end to end: extract data from email, update a spreadsheet, schedule a calendar event, and write a summary — all without switching contexts. For developers, the agent’s coding capabilities (write, debug, choose tools) could become a first-party alternative to GitHub Copilot or Cursor, embedded directly in Xcode and the broader Apple toolchain. The listing’s emphasis on coordinating sub-agents also hints at a platform where specialized agents handle domain-specific steps while a planner orchestrates the whole. Until Apple announces a product, the job listing remains a hiring signal — but a detailed one that shows exactly where the company intends to compete.
If you’re building on Apple platforms, watch the macOS 27.3 permission model closely. The same sandboxing and visibility rules that will govern third-party agents will likely shape what Apple’s own agent can do at launch.
Sources
- Report: Apple job listings detail plans for autonomous AI agents
- Apple Is Building an AI Agent to Use Apps, Browse the Web, and Write Code
- Apple job listing points to an in-house AI agent that codes…
- Apple Is Working on AI Agent for Device Control, Apps, Browsing and Coding
- Apple Is Working on AI Agent for Device Control, Apps, Browsing and Coding
- Apple developing its own AI agent to rival OpenAI, Anthropic and Grok …
- Apple works on agent to counter OpenAI, Anthropic and Meta: report
- Apple AI Agent for Device Control: 2027 Release Expected