SIRI AGENT: Next year again?

Published · 8 min read

Apple documents App Intents and usage limits for Siri AI. Here’s what’s confirmed, what’s reported, and how to prepare your business for assistant-led actions.

SIRI AGENT:
Next year again?

Apple’s intent and schema approach signals how assistants may execute tasks across apps

The question is not whether Siri will become more capable. Apple’s public trail already shows a staged, developer-driven path: apps expose actions and data in a structured way, then system features like Apple Intelligence and Siri can discover and use them. That is a very specific model, and it matters because it suggests how assistants will act on behalf of customers and employees.

Separately, Bloomberg reporting says Apple wants to turn Siri into a “true AI agent,” with secondary coverage framing that shift as possible as early as next year. Those claims have strategic weight, but they sit in a different evidence tier than Apple’s own documentation. For business leaders, the useful move is to plan for assistants that can take actions across software, under permissions and usage limits, and to make your services “legible” to that kind of interaction.

What Apple Publicly Documents Today

App Intents As The Integration Rail

Apple’s developer documentation describes “App Intents” as a framework where a developer “express[es] your app’s actions and data in a structured way that makes them discoverable by Apple Intelligence.” Apple also states that this structure supports “deeper integration with system features,” and that after compilation “Apple Intelligence, Siri, and other system features” can “discover and use your intents, entities, and app enum types.”

This is a concrete design choice. It frames assistant capability as something that expands through explicit, structured exposure of actions and data, rather than through an open ended promise of autonomy. For organizations, it points to a practical reality: assistants can only act reliably when the actions and the objects they act on are well-defined.

A Concrete Example: Photos Domain Actions

Apple’s documentation for the “Photos” domain under App schema domains describes the “.photos” domain as defining app schemas that provide “a structured representation for common photo and video actions.” Apple adds that you can apply schemas in this domain “to make your app’s photo and video functionality available to Apple Intelligence and Siri.”

The same documentation includes specific actions and examples that illustrate the level of granularity Apple is describing. It lists setRotation with example phrases like “Rotate this photo clockwise.” and “Rotate this photo left.” It also shows crop as an intent (var crop: some AppSchemaIntent).

The key implication is not about photos. It is about how capability is packaged: named actions, defined inputs, and predictable outcomes. If assistants move toward multi-step execution, this kind of structure becomes the difference between a helpful flow and a brittle one.

Apple Intelligence And Siri AI: Availability And Limits

Apple Support’s “Get started with Apple Intelligence on your iPhone” states: “With iOS 27, Apple Intelligence is integrated across your apps and experiences.” The same page also warns that “Apple Intelligence is not available in all languages or regions. Usage limits may apply.”

Apple Support also lists several specific conditions and constraints relevant to assistant-led actions:

  • “Siri AI will not be initially available in the EU on iOS, iPadOS, and watchOS.”

  • “Siri AI is not available for users under 13.”

  • Apple says that “Certain Apple Intelligence features that rely on server-side models are subject to daily usage limits,” including “Siri AI,” “Intelligent Photo Editing Tools,” “Image Playground,” and “AFM 3 Cloud models in Shortcuts.”

  • Apple adds: “Expanded access to such features will be available for a fee in the future.”

Separately, Apple’s “Apple Intelligence Usage Terms” states that “Apple Intelligence Features may be subject to usage limits,” and that Apple “reserves the right to modify” those limits “at any time.” The same terms describe “Unreasonable Use” (including “automated” or “abusive” activity) as a basis for access being “restricted, throttled, suspended, or subject to additional limits.”

For business and operations teams, this is the under-discussed foundation. As assistants shift from answering to doing, usage limits and enforcement language become part of the environment. Any workflow that depends on large volumes of automated requests or repeated retries is more exposed to throttling risk than a workflow designed around fewer, more reliable, well-defined actions.

What Is Reported Versus Confirmed

The Bloomberg Claim, Kept In Its Lane

Bloomberg’s Power On newsletter page (dated “November 8, 2026 at 3:00 PM GMT+0”) carries the headline: “Apple Explores Turning Siri AI Into ChatGPT Dots-Like Agent; iOS 28 Shake-Up.” The available snippet includes the line: “The company wants to turn Siri into a true AI agent.”

Secondary coverage frames this as a rumor. MacRumors, in “Siri AI Rumored to Become ‘More Like an Agent’ as Early as Next Year,” attributes the idea to Gurman’s Power On newsletter and uses the wording “as early as next year.”

For readers, the decision point is straightforward: treat this as reporting about internal direction and timing expectations, and treat Apple’s developer and support documentation as the reliable guide to what the company is building toward in public. The staged structure Apple documents fits the broad direction implied by the reporting, without turning the report into a product commitment.

A Useful Way To Translate The Reporting Into Planning

Even without accepting any specific roadmap details, the combination of (1) App Intents and schema domains, (2) Apple’s explicit “usage limits” and “Unreasonable Use” language, and (3) the reporting that Apple wants a “true AI agent” creates a practical planning frame for organizations: assistants will only transact safely and repeatably when your service actions can be represented as clean, permissioned steps, and when those steps can survive rate limits and enforcement rules.

That is a different readiness problem than “write better prompts.” It is a systems problem: defining actions, defining entities, defining confirmations, and designing flows that do not depend on unlimited retries.

What Changes For Customer Acquisition And Operations

Discovery Can Become Delegation

If assistants can discover and execute app actions, the user’s behaviour can shift from browsing interfaces to delegating goals. When that happens, your marketing surface area expands beyond what your website says. The assistant needs to know what you offer and which actions map to a user’s intent.

In Apple’s framing, that knowledge is expressed structurally through “intents” and “entities,” then discovered by Apple Intelligence and Siri. For businesses, the deeper implication is second-order: the assistant’s reliability depends on how precisely you define the objects it can operate on. Ambiguity that feels manageable for a human caller can break an assistant-led flow because the assistant must choose a specific action with specific inputs.

Conversion Starts To Look Like A Workflow

In an assistant-led environment, conversion is less about persuasion at a single moment and more about executing a series of discrete steps: select the right service, pick an option, confirm details, then finalize. Apple’s developer model suggests those steps will be mediated through structured actions exposed by apps.

Teams that sell services can use this to audit their customer journey for “action definition.” Where are the steps that require interpretation, back-and-forth, or manual cleanup? Those are the steps that will need clearer rules, clearer categories, or better structured intake before assistants can execute them reliably.

Usage Limits Become A Design Constraint, Not A Footnote

Apple’s public statements that usage limits may apply, that some server-side features have daily usage limits, and that “Expanded access” may be paid, create a new constraint for any organization imagining assistant-heavy flows. High-friction designs that require repeated assistant retries, constant re-checking, or redundant steps create more requests and more exposure to throttling.

One practical trade-off follows: designing for fewer, higher-confidence actions can beat designing for maximum automation. It may mean adding a confirmation step or narrowing an action’s scope so it works consistently. The marketing and operations payoff is predictability, because predictable execution produces cleaner follow-through and fewer support exceptions.

What Organizations Should Examine Next

Inventory The Actions You Want A Customer Or Employee To Delegate

Start by listing the real actions that matter commercially, operationally, or for customer experience. Phrase them as verbs with clear objects, in the style Apple’s documentation implies: a discrete action on a defined entity. Keep it grounded in what your business already does well, and make sure each action has an unambiguous “done” state.

Define The Entities And The Minimum Inputs

Assistants can only execute actions consistently when the underlying objects are cleanly defined. Map each action to the entity it operates on, plus the minimum inputs required to complete it without guesswork. The goal is to reduce situations where the assistant would need to infer what the user meant from vague language.

Design A Rate-Limit Friendly Flow

Given Apple’s usage limit language and the “Unreasonable Use” concept that includes automated use, treat request volume as a constraint. Build flows that minimize retries and reduce backtracking. Where a step is likely to fail, include a clear human handoff point so the user can still complete the task without repeated automated attempts.

Choose A Single Owner For Assistant Readiness

Assistant readiness spans product, marketing, operations, and support. Assign a single accountable owner who can coordinate how actions are defined, how exceptions are handled, and how the customer experience stays coherent when tasks move from screens to delegated execution.

Bloomberg’s framing of a “true AI agent” makes for an attention-grabbing timeline question. Apple’s own documentation makes the more useful point: assistants gain real power through structured app actions, and their impact is shaped by availability rules and usage limits as much as by capability.

If you want help turning these shifts into practical, supervised execution across content, local visibility, lead capture, and workflow design, IN-STORE can support you as a monthly marketing partner.

Contact IN-STORE to discuss assistant readiness and next steps

Sources:

More articles from IN-STORE · This article as Markdown · RSS