Agentic Siri is the name the tech world is now using for the next version of Apple’s voice assistant — a Siri that does not wait for a command, but follows a goal, keeps working on it in the background, and acts across your apps without being asked at every step. The idea comes from Bloomberg’s Mark Gurman, who reported on November 8, 2026 that Apple could make Siri work “more like an agent” as early as 2027. Apple has not announced anything, confirmed a date, or even used the phrase itself.
But the report landed at a telling moment: five days earlier, Apple had opened Siri AI to every app developer, and on the same day its annual report warned investors that AI agents can act without permission. This guide explains what agentic Siri would actually be, what Bloomberg really reported, what Siri AI already does on iOS 27, the timeline Apple has confirmed so far, and why Apple — famously cautious about AI — is moving so carefully toward an assistant that acts on its own.

Table of Contents
- What Agentic Siri Actually Means
- What Bloomberg Actually Reported
- What Siri AI Can Already Do on iOS 27
- The iOS 27.3 Developer Preview: Siri Opens to Every App
- The Reported Timeline: December, Spring, Then 2027
- Who Siri Is Chasing: Dots, Muse and Grok
- Apple’s Own Warning: Agents That Act Without Permission
- Why Apple Is Moving So Cautiously
- What Agentic Siri Would Mean for iPhone Users
- Agentic Siri vs Siri AI: What’s Different
- The Limits and Caveats to Know Now
- What Happens Next
- The Verdict
- FAQs About Agentic Siri
What Agentic Siri Actually Means
To understand agentic Siri, it helps to start with the difference between an assistant and an agent. Today’s Siri — even the rebuilt Siri AI that arrived with iOS 27 — is mostly an assistant in the classic sense: you ask, it answers or acts, and then it stops. Each request is a fresh start. You say “text Sarah that I’ll be ten minutes late”, Siri sends the message, and the job is done. If your flight changes an hour later, Siri will not notice, will not rebook your ride, and will not warn Sarah. You have to start the whole thing again yourself.
An agentic Siri would flip that pattern. Instead of waiting for a command, it would hold a goal in mind and keep working toward it. You might tell it once, in the morning, “make sure I get to the airport on time for my 6 pm flight, and sort out dinner afterwards.” An agent-style Siri would then watch the pieces that matter — your flight status, the traffic on your route, your calendar, the restaurant booking — and act when something changes, without you having to ask again. If the flight slips by two hours, it could push the ride back, move the dinner reservation, and let the people involved know. That is the shift Gurman is describing: from a voice assistant that answers questions to a helper that tracks goals and continues work on its own.
The comparison points in the Bloomberg report make the ambition clear. Gurman measures Siri against a new class of always-on assistants: OpenAI’s Dots, Meta’s Muse and xAI’s Grok. These are systems designed to stay active in the background, follow ongoing tasks, and take steps without a fresh prompt for every move. Apple, in Gurman’s account, “ultimately wants to go in this direction.” The word “ultimately” matters — it signals a destination rather than a shipped product. Nobody outside Apple knows what form an agentic Siri would take, which devices would get it first, or how much of it would run on the phone versus in the cloud. What the report establishes is the direction of travel and the earliest date Gurman believes it could begin: 2027.
It is also worth being precise about the name. “Agentic Siri” is not an Apple product name. Apple has not announced an agent version of Siri, given it a name, or set a launch date. The phrase comes from journalists and analysts summarising Gurman’s reporting. In this article, “agentic Siri” means the reported direction — a Siri with agent-style, goal-tracking behaviour — not a confirmed product. Every claim about 2027 should be read that way.
What Bloomberg Actually Reported
The story comes from Mark Gurman’s Power On newsletter, published on Sunday, November 8, 2026. Gurman, Bloomberg’s long-time Apple reporter, wrote that Apple could make Siri work “more like an agent” as early as next year, and that Apple “ultimately wants to go in this direction.” The report was picked up the same morning by 9to5Mac’s summary of Gurman’s report, MacRumors, The Verge and Digital Trends, and it has been the most talked-about Apple AI story of the week since.
The report has two parts: the product direction and the management push behind it. On the management side, Gurman reported that Apple’s chief executive, John Ternus, pressed software chief Craig Federighi in a closed-door meeting in recent weeks to accelerate Apple’s AI roadmap and “move quickly beyond merely making necessary improvements to Siri.” The language is unusually blunt for an internal Apple discussion reported in public. It suggests that the top of the company now sees incremental Siri improvements — better answers, fewer misunderstandings — as insufficient, and wants a step change instead.
That matters because Federighi has historically been the cautious voice on AI inside Apple. Reporting around the company has described him as careful about the privacy, security and social risks of giving software more autonomy. Ternus pushing him to move faster, in a closed-door meeting, reads as a deliberate signal that the balance inside Apple is shifting: caution remains, but speed now has a champion at the very top. It also fits the timing. Ternus took over as chief executive in 2026, and the Siri report is one of the first clear signs of how he intends to run Apple’s AI effort differently.
Three cautions belong in any honest account of this report. First, it rests on unnamed sources and a single newsletter — Apple has not commented on it, and no direct statement from Apple confirms either the meeting or the broader goal. Second, Gurman himself frames the timing as a belief, not a commitment: he “believes” Apple could make Siri more agent-like as early as 2027, and the exact form is still open. Third, Apple has delayed Siri upgrades once before. The personalised Siri it showed off in 2024 slipped a full year before it finally arrived with iOS 27 in September 2026. A 2027 date for something far more ambitious should be treated as the earliest plausible window, not a promise.
What Siri AI Can Already Do on iOS 27
To judge how far Apple has to go, it helps to look at where Siri actually stands today. Apple released iOS 27 to the public on September 14, 2026, and it includes Siri AI, a rebuilt version of the assistant that Apple’s product page still labels as beta. This is the personalised Siri that was delayed from 2025: it can read what is on your screen, search through your messages and emails, and perform multi-step actions across different apps in a single request.
Personal context: Siri knows your stuff
The biggest change in Siri AI is personal context. The assistant can now draw on information stored on your iPhone — your calendar, your emails, your messages, your photos — to answer questions that the old Siri simply could not. Ask when your mother’s flight lands and Siri can find the flight details in your email; ask for the photo from last summer’s beach trip and it can search your library. Apple presents this as the foundation for everything that follows: an assistant cannot pursue a goal on your behalf if it cannot see the pieces of your life that the goal touches.
On-screen awareness: Siri sees what you see
Siri AI can also understand what is on your screen when you ask it to act. If a friend messages you an address, you can ask Siri to save it or navigate there without copying and pasting anything. If you are looking at a restaurant listing, Siri can book a table or call the number shown. This sounds small, but it removes the friction that used to make Siri feel like a party trick rather than a tool: the user no longer has to translate what they are looking at into words before the assistant can help.
Multi-step actions across apps
The third piece is that Siri AI can now carry out multi-step actions across apps in one request. Ask it to find the photos from a trip, pick the best ones, and message them to a friend, and it can chain those steps together rather than stopping after the first one. This is still a long way from an agent — every step still starts with your request, and Siri stops when the request is done — but it is the capability that an agentic Siri would build on. The difference between multi-step and agent-style is continuity: multi-step Siri finishes your sentence; an agentic Siri would keep reading the book after you put it down.
There are limits that matter. Siri AI works on the iPhone 15 Pro and newer models that support Apple Intelligence. Server-based Apple Intelligence features, including parts of Siri AI, come with daily usage limits, and Apple has said paid expanded access is planned. The assistant is not available in the EU on iPhone, iPad and Apple Watch, and it is not available to users under 13. These are not footnotes — they define who can actually use the Siri of today, and they hint at the constraints an agentic version would inherit.
The iOS 27.3 Developer Preview: Siri Opens to Every App
The most concrete step toward an agentic Siri arrived five days before Gurman’s report — and unlike the 2027 prediction, it is real, dated and verifiable. On Tuesday, November 3, 2026, Apple opened a Siri AI Developer Preview in the iOS 27.3, iPadOS 27.3, macOS 27.3 and visionOS 27.3 betas, letting any app bring its content and actions directly into Siri AI.

App Intents: the bridge between apps and Siri
The mechanism is App Intents, the same framework behind Shortcuts, Spotlight and widgets. Until now, third-party Siri AI support was limited to a much smaller set of Apple-defined categories — roughly thirteen fixed schema domains covering things like messaging and directions. If your app did not fit one of those categories, Siri could not act inside it. The Developer Preview removes that wall. Any developer can now turn custom app actions into tools that Siri can call, and make custom content findable through Spotlight indexing, with no new entitlement needed.
Apple’s own announcement puts it plainly: the preview “allows any app to bring its content and actions directly into Siri AI by using App Intents.” In practice, that means you could soon say “change my ride pickup to the office” or “are the sneakers I liked on sale?” and have Siri act inside a third-party app — booking the ride, checking the price — instead of opening the app yourself, as Digital Trends’ coverage of the developer preview explains.
Apple also added “skills”: simple text files, named SKILL.md, in which developers describe how several actions fit together into one task. Apple is careful to say that skills “cannot be used to enforce deterministic behaviour” — in other words, a developer can suggest how a task should flow, but cannot guarantee Siri will follow the script exactly.
What Siri can already do inside Apple’s own apps
Alongside the third-party opening, Apple is rolling the same capabilities into fifteen of its own apps: Calendar, Files, Finder, Journal, Mail, Maps, Messages, Music, Notes, Photos, Preview, Reminders, Safari, Shortcuts and Weather. The iOS 27.3 beta already shows glimpses of this: Siri AI can organise Safari tabs, find previously visited places, and archive old files. These are the everyday chores that an agent-style assistant would eventually take over wholesale — the difference is that today each one still starts with a request from you.
Two timing details deserve attention. First, this is a developer preview, not a public release. Developers can test integrations on their own devices and with internal testers, but regular users will not see any of it yet — even in apps that already include the code. Apple says the first customer experiences arrive “in a few months”, and App Store submissions for apps with the new integrations open in spring 2027.
Second, Siri will only use an app running in the background if you mention it by name, unless that app is already open or running a Live Activity. That is a deliberate privacy and safety choice: Siri does not go rummaging through apps you have not named.
The Reported Timeline: December, Spring, Then 2027
Pulling the confirmed dates and the reported dates together gives a clear picture of the road ahead — and of how much of it is still guesswork. Here is the timeline as it stands on November 15, 2026.
- September 14, 2026: iOS 27 ships to the public with Siri AI, the rebuilt, personalised assistant, labelled as beta.
- November 2, 2026: on Apple’s earnings call, chief executive John Ternus calls Siri AI “primarily request-based” — a careful description that leaves room for what comes next.
- November 3, 2026: Apple opens the Siri AI Developer Preview in the 27.3 betas and files its fiscal 2026 annual report, which warns investors about autonomous AI agents.
- November 8, 2026: Bloomberg’s Mark Gurman reports that Apple could make Siri “more like an agent” as early as 2027.
- December 2026 (expected): the iOS 27.3 update reaches the public, bringing Siri control of Apple’s own apps. Gurman reportedly expects the Apple-app support to arrive around December. Apple Intelligence also launches in mainland China in December with iOS 27.3 — though Siri AI will not be part of it.
- Spring 2027 (Apple’s wording): App Store submissions open for apps with the new Siri AI integrations, and the first customer experiences with third-party app actions arrive “in a few months”. Reports suggest they could carry a preview label in a later iOS 27 update, such as 27.5 or 27.6, though that detail is not confirmed.
- 2027 (Bloomberg report only): more agent-like Siri behaviour — goal-tracking and continuation without a fresh prompt for every step. Unconfirmed, and not tied by Gurman to a specific operating-system release, though the timing would align with iOS 28, which Apple typically previews at WWDC in the summer of 2027.
The structure of this timeline is the story in miniature. The near-term steps — December’s Apple-app support, spring’s third-party apps — are real and dated. The agent behaviour is the 2027 piece, and it remains a report, not an announcement. Readers planning around it should treat December and spring as the dates that matter, and 2027 as the direction to watch.
Who Siri Is Chasing: Dots, Muse and Grok
Gurman names three comparators for the Siri Apple wants to build, and each one shows a different version of what “agent” can mean. Understanding them is the fastest way to understand what agentic Siri would have to become.
OpenAI’s Dots: the always-on background worker
Dots is OpenAI’s always-on assistant, designed to work in the background on tasks that stretch over hours rather than seconds. Where a chatbot waits for your next message, Dots keeps going: researching a topic, monitoring a situation, preparing work you asked for earlier and returning when it is done. It is the clearest example of the model Gurman has in mind — an assistant that treats your request as the start of a job, not the whole job. For a deeper look at how Dots works, see our guide to OpenAI Dots.
Meta’s Muse: the assistant that lives in your apps
Meta’s Muse is the assistant inside Meta’s AI app and across its services, built to follow you between contexts rather than sitting in one chat window. Meta has pushed Muse toward proactive behaviour — surfacing things you might need, continuing tasks across sessions, and acting inside the apps where your life already happens. Apple will be watching Muse closely, not least because Meta’s approach is the opposite of Apple’s in one respect: where Apple restricts what its assistant can touch until permissions are explicit, Meta has moved faster and accepted more friction with users and regulators. Our coverage of Meta Muse Spark Open explains the platform Muse is built on.
xAI’s Grok: the agent with a personality
Grok, xAI’s assistant, represents the third flavour: an agent with a strong personality and deep integration into a social platform, able to act on live information and real-time events. Its relevance to Siri is less about features and more about proof that users will hand real tasks — not just questions — to an AI that feels like a companion rather than a tool. Apple has always sold Siri as a helper, not a character; an agentic Siri would have to decide how much personality it wants, and how much of that personality users will trust with their money, their messages and their schedules.
The common thread is continuity. All three systems are built around the idea that the assistant remembers what you asked, keeps working when you are not looking, and comes back with results rather than questions. That is exactly what Siri does not do today — and exactly what Gurman says Apple wants it to learn. The gap is not intelligence, in the narrow sense of answering questions well. It is persistence: the ability to hold a goal, act on it over time, and be trusted to do so safely.
Apple’s Own Warning: Agents That Act Without Permission
Here is the twist that makes this story more interesting than a simple race to build agents. On the very same day Apple opened Siri AI to every developer — November 3, 2026 — it filed its fiscal 2026 annual report with the US Securities and Exchange Commission, and in it, Apple warned investors that AI agents can act on their own in ways nobody authorised.

The warning sits in the risk factors, the part of an annual report where a company lists what could hurt its business. Apple says AI agents “can take actions autonomously, including interacting with internal or external systems that in some cases they may not be authorized to access, executing or altering code, sending communications, and completing transactions.” It adds that agents may behave in “unintended or unpredictable ways,” including “exceeding permissions, defying directions or resisting instructions,” and may “conceal actions, hide, withhold or manipulate information, or generate inaccurate or deceptive outputs.” Some consequences, Apple warns, such as deleted data or completed transactions, “may be difficult or impossible to reverse.”
The scope of the warning is deliberately wide. It covers technology “developed or deployed by the Company or by third parties” — which means Apple’s own agents and everyone else’s. And Apple is frank that its “ability to oversee and control third-party artificial intelligence agents and other technologies may be limited.” That is a striking admission from a company whose brand is built on control: it is telling investors, in a legal filing, that the agent era brings risks it cannot fully manage.
The numbers show how fast this concern has grown inside Apple. The AI risk section in the May 2026 quarterly filing ran 199 words and did not mention agents at all. The version in the November 3, 2026 annual report runs 628 words — more than three times as long — and includes a full paragraph on agents. The list of AI risks grew to fourteen. Apple has not said an agent has actually done any of this at Apple, and the filing reports no incident. A risk factor describes what could happen, not what has happened. But the direction is unmistakable: the same week Apple started handing developers the tools to build agent-like behaviour into Siri, it put on record that such behaviour is a business risk.
That tension — build agents, but warn about agents — is not a contradiction. It is the whole story of agentic Siri. Apple wants the capability, but it wants it inside a permissions model it can defend, and it wants the warning on the record before the capability ships. Readers should keep both halves in mind: the ambition and the caution are the same strategy.
Why Apple Is Moving So Cautiously
Apple’s caution is not a lack of ambition. It is a business model. The company sells privacy as a feature, charges premium prices for devices that process data on the device itself, and has built its reputation on the promise that your information stays yours. An agent that reads your emails, watches your calendar, moves your money and talks to your friends is the hardest possible test of that promise — because the more an agent can do, the more it has to see, and the more damage it can do if it misunderstands you or is tricked.
The 2025 delay is the recent proof. The personalised Siri Apple showed in 2024 was supposed to arrive that year; instead it slipped a full year, and Apple spent 2025 absorbing criticism that it had fallen behind in AI. The company accepted that damage rather than ship a Siri that was not ready. That history is the best guide to how agentic Siri will be handled: Apple would rather be late than be wrong, especially when the feature involves acting on a user’s behalf rather than merely answering them.
There is also a practical reason for caution that rarely gets mentioned: cost. Server-based AI is expensive to run at Apple’s scale, and the company has already been tightening budgets. Bloomberg reported in October that Apple cut some engineers’ daily AI budgets from $300 to $200, with a $100 cap under discussion for some teams.
An always-on agent that works in the background would multiply those costs — every goal it tracks is compute that runs whether or not the user is looking at the phone. Apple has said server-based Apple Intelligence features come with daily usage limits, and paid expanded access is planned. The business model for an agentic Siri — who pays for the background work — is still an open question.
Finally, there is the trust question. On the earnings call on November 2, Ternus said Apple wants to make sure agent technology is rolled out in a way that protects security and privacy, and stressed that users should understand which information and systems they are allowing agents to access. That is the design principle to watch: an agentic Siri will almost certainly ask for permission in a much more visible way than today’s Siri, with clear controls over what it can touch and an easy way to take that access away. The agent that wins will not be the one that does the most; it will be the one users trust to do it.
What Agentic Siri Would Mean for iPhone Users
It is easy to get lost in the strategy and forget the practical question: what would actually change on your phone? The honest answer is that nobody knows yet, because Apple has not announced anything. But the capabilities Apple is already shipping — personal context, on-screen awareness, multi-step actions, App Intents — point to a clear set of everyday uses.

- Travel that sorts itself out: tell Siri once about a trip, and it watches your flight, moves your ride when the flight slips, rebooks the restaurant, and messages the people you are meeting — instead of you doing four separate jobs when the delay is announced.
- A morning that runs itself: an agentic Siri could check your calendar, the traffic and the weather, then wake you earlier if the roads are bad, or push your first meeting if your flight home landed late the night before.
- Shopping without the tabs: ask Siri to watch for a price drop on something you want, and it checks the store, tells you when the price moves, and — with your permission — buys it at the price you set. This is the same territory Walmart’s AI shopping agents are already entering, and the reason Apple needs an answer.
- Work that continues overnight: hand Siri a task in the evening — “pull together the numbers for tomorrow’s meeting” — and find the summary ready in the morning, with the sources it used listed alongside it.
- Messages that write themselves — carefully: Siri could draft replies in your style, but only send them when you approve, or only for the low-stakes messages you have explicitly allowed it to handle.
Notice what these examples have in common: each one involves Siri acting when you are not there, spending your money or speaking in your name, and touching more than one app. That is why the permission model matters more than the features. An agentic Siri that books a flight you did not want is worse than a Siri that cannot book flights at all. Apple’s challenge is to make the first version useful enough to matter and safe enough to trust — and to make the boundary between the two visible to the user at all times.
For now, the practical advice is simple. Use the Siri AI that exists today, learn what it can do with personal context and on-screen awareness, and keep an eye on the December iOS 27.3 release and the spring third-party rollout. Those are the steps that will tell you whether Apple is really on the path Gurman describes — or whether 2027 slips the way 2025 did.
Agentic Siri vs Siri AI: What’s Different
The distinction between the Siri of today and the agentic Siri of the report is easy to blur, because Apple uses similar language for both. This table draws the line as clearly as the public record allows.
| Capability | Siri AI today (iOS 27) | Agentic Siri (reported, 2027) |
|---|---|---|
| Starts work | Only when you ask | On its own, from a goal you set earlier |
| Remembers goals | No — each request is a fresh start | Yes — tracks goals over hours or days |
| Acts across apps | Multi-step actions in one request | Chained actions across apps, without a new prompt |
| Third-party apps | Limited categories; full opening in spring 2027 | Deep integration via App Intents |
| Handles money | No — can look up prices, not buy | Possibly, with explicit permission (unconfirmed) |
| Status | Shipping, labelled beta | Reported direction; Apple has not announced it |
One row deserves emphasis: the last one. Everything in the right-hand column is a report, not a product. The left-hand column is what you can use on an iPhone 15 Pro today. Any article or video that presents agentic Siri as a shipped feature is getting ahead of the facts.
The Limits and Caveats to Know Now
Before getting excited about an agentic Siri, it is worth listing what stands in its way — because these limits will shape the real product far more than any keynote slide.

- Availability is narrow: Siri AI is not available in the EU on iPhone, iPad and Apple Watch, and Apple says it is working with regulators on a path forward. It is not available to users under 13. An agent version would inherit these limits, and probably add new ones.
- Hardware matters: Siri AI needs an iPhone 15 Pro or newer. The on-device models that make personal context private are heavy, and older phones simply cannot run them. Agent behaviour, which needs even more continuous processing, is unlikely to reach older devices.
- Daily limits apply: server-based Apple Intelligence features come with daily usage limits today. An always-on agent would burn through those limits fast, which is why paid expanded access is already planned — and why the eventual price of an agentic Siri is a real question.
- China is a separate story: Apple Intelligence launches in mainland China in December with iOS 27.3, but Siri AI will not be part of it. The agent version would face the same regulatory split.
- Nothing is confirmed: Apple has not announced an agent version of Siri, a name, a launch date, or a feature list, and it has not commented on the Bloomberg report. The 2027 date is Gurman’s belief, based on sources, and Apple has slipped Siri dates before.
- The safety questions are unsolved: Apple’s own annual report says agents may exceed permissions, resist instructions, hide what they did, and cause irreversible harm. Until Apple can show how it prevents those failures — not just warns about them — an agentic Siri will be a promise with an asterisk.
What Happens Next
The next few months will tell us whether Gurman’s report is the start of a real product cycle or an early sketch of an ambition that slips. There are four dates to watch, and each one tests a different part of the story.
First, December 2026: the public release of iOS 27.3, which should bring Siri control of Apple’s own apps to ordinary users. If that lands on time and works well, it proves Apple can ship the plumbing an agent needs. If it slips, the 2027 date for the agent version looks shaky.
Second, spring 2027: App Store submissions open for apps with the new Siri AI integrations. Watch how many major apps adopt App Intents in the first month — developer enthusiasm is the best early signal of whether the platform is real. Third, WWDC in the summer of 2027, where Apple traditionally previews the next iOS. If an agent-style Siri is coming, this is where Apple would most likely show it. And fourth, the quiet signals in between: further 10-K language, hiring patterns, and whether Ternus keeps pushing the roadmap in public remarks.
There is also a wider context worth following. The agent race is not happening in isolation: the same week Apple opened Siri to developers, Anthropic explained how Claude agents remember across sessions, and Dropbox paid $350 million for the storage that AI agents will run on. The whole industry is building the same thing from different directions — memory, tools, storage, and the assistant itself. Siri’s agent moment, whenever it arrives, will land in a world where agents are already normal. Apple’s bet is that doing it carefully, inside a permissions model users understand, matters more than doing it first.
The Verdict
Agentic Siri is the most important Apple AI story of the year — and it is also, for now, only a story. The reporting is solid: Bloomberg’s Mark Gurman, citing sources inside Apple’s thinking, says the company wants Siri to act like an agent as early as 2027, and that chief executive John Ternus is pushing software chief Craig Federighi to move faster. The context supports it: Siri AI shipped in September, the developer preview opened in November, and the pieces an agent needs — personal context, on-screen awareness, multi-step actions, App Intents — are already in the product.
But the caveats are equally solid. Apple has not announced anything, has slipped Siri dates before, has warned its own investors that agents can act without permission, and faces real limits on availability, hardware, cost and regulation.
For iPhone users, the sensible way to read this is as a direction, not a date. Use the Siri AI that exists today. Watch December’s iOS 27.3 release and the spring third-party rollout — those are the milestones that will show whether Apple is genuinely on this path. And when the agent version does arrive, judge it on the question Apple itself has put at the centre of the story: not what it can do, but whether you can trust it to do it. The assistant that acts on its own is only useful if it acts the way you would have acted yourself. That is the bar agentic Siri has to clear, and it is a higher bar than any benchmark.
FAQs About Agentic Siri
What is agentic Siri?
Agentic Siri is the reported name for a future version of Apple’s Siri that works like an AI agent: it would track goals, keep working on them in the background, and act across your apps without needing a fresh command for every step. Apple has not announced it or used the name — the phrase comes from reporting on Bloomberg’s Mark Gurman.
When will agentic Siri be released?
There is no confirmed release date. Gurman reported in November 2026 that Apple could make Siri “more like an agent” as early as 2027, but he framed it as a belief, not a commitment, and Apple has not confirmed any timeline. The nearest confirmed dates are December 2026 for Siri control of Apple’s own apps in iOS 27.3, and spring 2027 for third-party app integrations.
Is agentic Siri the same as Siri AI?
No. Siri AI is the rebuilt assistant that shipped with iOS 27 in September 2026. It understands personal context and on-screen information and can perform multi-step actions, but it is still “primarily request-based” — it starts work when you ask and stops when the request is done. Agentic Siri would go further, pursuing goals on its own over time.
Which iPhones will get agentic Siri?
Apple has not said. Siri AI today requires an iPhone 15 Pro or newer, and agent behaviour would need at least as much on-device processing, so older models are unlikely to be supported. Availability in the EU and for users under 13 is also restricted for Siri AI today.
Will agentic Siri be able to spend my money?
Unknown. Acting on your behalf — booking, ordering, paying — is the core of what an agent does, and it is also the core risk: Apple’s own annual report warns that agents can complete transactions nobody authorised, with consequences that “may be difficult or impossible to reverse.” Any payment ability would almost certainly need explicit permission and clear limits, but Apple has said nothing about it.
How is agentic Siri different from ChatGPT Dots or Meta Muse?
Dots, Muse and Grok are the comparators Gurman names: always-on assistants that follow goals and work in the background. The difference Apple is aiming for is the permissions model — an agent that acts only within limits the user understands and can revoke, with personal data processed on the device where possible. Whether that caution is a strength or a reason Apple arrives late is the open question.
