Apple Intelligence Agents Come to iOS 20: Siri Can Now Book Flights, Order Food, and Manage Email

Siri Finally Becomes the Assistant Apple Always Promised

At a special event held at Apple Park on March 12, 2026, Apple unveiled the most significant upgrade to Siri in the assistant's 15-year history. Branded as "Apple Intelligence Agents," the new capabilities transform Siri from a voice-activated search and command tool into a genuine autonomous agent capable of completing complex, multi-step tasks across apps without constant user supervision.

Craig Federighi, Apple's Senior VP of Software Engineering, demonstrated Siri booking a round-trip flight from San Francisco to Tokyo by searching multiple airlines, comparing prices and schedules against the user's calendar availability, selecting optimal flights, filling in passenger information from the user's profile, and completing the purchase through Apple Pay, all from a single natural language request.

"For years, Siri has been able to answer questions and set timers," Federighi said. "Starting with iOS 20, Siri can actually do things for you, real things that previously required 15 minutes of tapping through multiple apps."

What Apple Intelligence Agents Can Do

Apple demonstrated seven core agent capabilities that will ship with iOS 20 this fall:

flowchart LR
    INPUT(["User intent"])
    PARSE["Parse plus<br/>classify"]
    PLAN["Plan and tool<br/>selection"]
    AGENT["Agent loop<br/>LLM plus tools"]
    GUARD{"Guardrails<br/>and policy"}
    EXEC["Execute and<br/>verify result"]
    OBS[("Trace and metrics")]
    OUT(["Outcome plus<br/>next action"])
    INPUT --> PARSE --> PLAN --> AGENT --> GUARD
    GUARD -->|Pass| EXEC --> OUT
    GUARD -->|Fail| AGENT
    AGENT --> OBS
    style AGENT fill:#4f46e5,stroke:#4338ca,color:#fff
    style GUARD fill:#f59e0b,stroke:#d97706,color:#1f2937
    style OBS fill:#ede9fe,stroke:#7c3aed,color:#1e1b4b
    style OUT fill:#059669,stroke:#047857,color:#fff

Travel Booking Agent

Siri can search flights, hotels, and rental cars across multiple providers, compare options based on user preferences (learned from past behavior), check calendar conflicts, and complete bookings through Apple Pay. The agent handles the full workflow including booking confirmation emails and calendar event creation.

Food and Grocery Agent

Users can ask Siri to "order my usual from Thai Palace" or "get groceries for the pasta recipe I saved last week." The agent integrates with delivery apps (DoorDash, Uber Eats, Instacart) and restaurant ordering systems, handling menu selection, customization, payment, and delivery tracking.

Email Management Agent

Perhaps the most immediately useful capability, the email agent can read, categorize, prioritize, draft responses, and schedule sends. Federighi demonstrated asking Siri to "go through my inbox, flag anything urgent, draft responses to the routine ones, and put them in my drafts for review." The agent processed 47 emails in under two minutes, generating contextually appropriate draft responses that the user could review, edit, and send with a single tap.

Calendar Optimization Agent

The agent can reorganize schedules, find meeting times across participants, suggest rescheduling when conflicts arise, and proactively block focus time based on the user's work patterns. It integrates with both Apple Calendar and third-party calendar services.

Health and Fitness Agent

Using data from Apple Health, Apple Watch, and connected fitness apps, the agent can create and adjust workout plans, schedule wellness activities, and even book medical appointments through participating healthcare provider systems.

Home Automation Agent

Building on HomeKit, the agent manages smart home devices in a contextual way. Rather than simple commands like "turn off the lights," users can say "I'm going to bed" and the agent will execute a personalized sequence: locking doors, adjusting thermostats, enabling security cameras, dimming lights, and setting morning alarms.

Research Agent

Users can ask Siri to research a topic and compile findings. The agent browses the web, reads articles, extracts key information, and produces a structured summary with sources. This capability is powered by Apple's new "WebKit AI" engine that can parse and understand web content in real time.

The Technical Architecture: On-Device Meets Cloud

Apple's approach to agentic AI reflects its longstanding commitment to privacy. The system uses a hybrid architecture:

On-device processing handles task planning, personal context retrieval (calendar, contacts, preferences, health data), and simple agent actions. Apple's custom A20 chip includes a dedicated "Agent Neural Engine" with 40 TOPS (trillion operations per second) of performance specifically optimized for agent workloads.

Private Cloud Compute handles more complex reasoning, web interactions, and multi-step planning that exceeds on-device capabilities. Apple emphasized that Private Cloud Compute runs on Apple Silicon servers, data is processed in encrypted enclaves, no user data is stored after processing, and the system has been audited by independent security firms.

The App Intents framework has been massively expanded. Third-party developers can now expose not just simple actions but complex workflows to Siri agents. Apple announced that over 2,000 apps have already integrated with the new App Intents API during the developer preview period, including major apps from Airbnb, United Airlines, DoorDash, Slack, and Notion.

The action verification system is a notable safety feature. Before executing any action that involves spending money, sending communications, or modifying data, Siri presents a clear summary of what it intends to do and asks for confirmation. Users can configure trust levels on a per-action-type basis: always confirm for purchases over $50 but auto-approve for calendar changes, for example.

How It Compares to Google and Samsung

Google's Gemini assistant, which received major agentic upgrades in January 2026, offers similar multi-step task completion. However, Google's approach is more cloud-dependent, sending more data to Google's servers for processing. Google's advantage is its deeper integration with web services and search.

Samsung's Galaxy AI, powered by a combination of on-device models and Google's Gemini, offers agentic features but primarily within Samsung's own app ecosystem. Cross-app agent workflows are more limited compared to Apple's App Intents approach.

The key differentiator for Apple is the privacy architecture. In a Pew Research survey conducted in February 2026, 67% of respondents said they would be more comfortable using an AI agent that processes data on-device rather than in the cloud. Apple's hybrid approach directly targets this preference.

Developer and Analyst Reactions

The developer community response has been overwhelmingly positive, particularly regarding the expanded App Intents framework. David Smith, an independent iOS developer, wrote that "Apple Intelligence Agents are what SwiftUI was for UI development, a paradigm shift that will redefine how we think about app functionality."

Wall Street analysts were similarly enthusiastic. Morgan Stanley's Erik Woodring raised his Apple price target, writing that "agentic AI transforms Siri from Apple's biggest weakness into potentially its greatest strength" and estimated the feature could drive a significant iPhone upgrade cycle.

Not all reactions were positive. Privacy advocates, while praising the on-device approach, expressed concern about the food ordering and travel booking capabilities, which necessarily involve sharing personal data with third-party services. The Electronic Frontier Foundation (EFF) published a statement urging Apple to provide granular controls over what data agents can share with external services.

Consumer advocacy groups raised concerns about the purchasing capabilities. "An AI agent that can spend your money introduces new categories of risk," said Consumer Reports' director of digital policy. "What happens when Siri misunderstands a request and books the wrong flight? Apple needs clear policies on agent-initiated purchase disputes."

The Competitive Implications

Apple's entry into agentic AI has immediate implications for the broader industry. With over 1.5 billion active Apple devices worldwide, iOS 20's agent capabilities will likely expose more consumers to agentic AI than any previous product launch.

This creates pressure on the entire value chain. Standalone AI assistant startups, many of which have raised hundreds of millions in venture funding, now face competition from a pre-installed feature on every iPhone. App developers must decide whether to deeply integrate with Apple's App Intents framework or risk being bypassed by Siri agents that can only interact with cooperating apps.

For enterprise IT departments, the question is whether Apple Intelligence Agents will enter the workplace in the same way the iPhone itself did, through employee demand rather than top-down deployment. Early signs suggest yes: several Fortune 500 CIOs interviewed by Bloomberg said they are already developing policies for AI agent use on corporate devices.

When It Ships

iOS 20 with Apple Intelligence Agents will enter developer beta in June 2026 at WWDC, with a public beta in July and general availability in September 2026. The full agent feature set requires an iPhone 16 or later (for on-device processing) or any iPhone with an active iCloud+ subscription (for cloud-based processing with reduced on-device capabilities).

Sources

• Apple Newsroom, "Apple Introduces Apple Intelligence Agents for iOS 20," March 2026
• The Verge, "Siri can finally do things: Apple Intelligence Agents are the biggest upgrade in years," March 2026
• Bloomberg, "Apple's Agentic AI Play Targets Google and Samsung in the Assistant Wars," March 2026
• TechCrunch, "iOS 20 turns Siri into a real AI agent that can book flights and manage your email," March 2026
• Wired, "Apple's Privacy-First Approach to AI Agents Could Be a Winning Strategy," March 2026