Gemini task automation reaches more than 40 apps

Definition: Gemini Intelligence is Google’s push to make Gemini act across apps and devices, not just answer a prompt in one app.
What changed: Task automation has expanded from a handful of beta apps to more than 40 popular apps on Samsung’s newest foldables.
Hardware direction: Google and Samsung also showed two more Android XR eyewear designs from Gentle Monster and Warby Parker.
Key takeaway: The announcement connects agentic phone actions with a wearable interface that can understand visual context while keeping the user’s hands and attention freer.

Google’s Gemini strategy is moving from “ask an assistant” to “delegate a chore.” At Galaxy Unpacked, the company announced that Gemini task automation now supports more than 40 popular apps, up from the handful of rideshare and delivery apps available in its earlier beta.

The same event put a second piece of the strategy on display: Samsung and Google showed two additional Android XR eyewear designs from Gentle Monster and Warby Parker, ahead of a planned fall launch.

Those announcements belong together. One expands what Gemini can do across a phone. The other changes where the user can ask for help. The long-term product is not a chatbot trapped inside an app. It is a cross-device agent that can see context, take a multi-step action, pause for review and continue when the user approves.

How Gemini task automation expands across supported apps

Google’s Galaxy Unpacked announcement says task automation is expanding from a handful of apps to more than 40 popular apps. The company names shopping, dinner reservations, travel experiences and event tickets as the broader categories now in scope.

The earlier beta focused on simple chores such as ordering a rideshare or getting takeout. The new direction is closer to “life admin”: a user describes the outcome, Gemini navigates the supported service and the user steps in when a decision or final confirmation is required.

The difference is important:

The original report notes Etsy and Expedia among the examples discussed around the rollout. Google’s own announcement emphasizes the categories and says availability includes supported apps across the US and Korea, with device, market, language and age limitations.

That qualifier is not a footnote to ignore. “40-plus apps” describes the supported surface area, not universal automation across every app a user has installed.

How the Gemini automation loop works

A useful way to understand Gemini’s new behavior is as a controlled agent loop:

1. The user states an outcome, such as finding dinner or booking an experience.
2. Gemini interprets the request and navigates the relevant supported app.
3. The user watches step-by-step progress on screen or through notifications.
4. The user can multitask, interrupt, adjust a detail or take over briefly.
5. Gemini can continue the remaining work in the background.
6. The user receives a prompt to review and confirm the finished task.

This is the practical territory of an AI agent: a model interprets a goal, uses tools, maintains state across steps and returns control when the action has consequences.

The human checkpoint is central. A system that can place an order, buy a ticket or make a reservation should not silently treat an ambiguous intermediate result as permission to finish. Google says Gemini can stop when the task is ready for review and final confirmation.

That makes the product feel less like “autopilot” and more like delegated execution with a visible approval boundary.

How Galaxy Z Flip8 changes Gemini access

On the Galaxy Z Flip8, Google says users can access task automation directly from the Flex Window by long-pressing the power button. The shortcut matters because it reduces the friction between noticing a chore and delegating it.

A phone assistant normally competes with the app the user is already looking at. A Flex Window shortcut makes Gemini a device-level action surface instead of a destination that must be opened first.

That does not automatically make the interaction better. The shortcut must still make it clear what Gemini understood, which app it will use, what information it will share and when the user is expected to confirm. The lower the friction, the more important those boundaries become.

How Gemini uses screen and image context

Google also says Gemini can use advanced reasoning to understand what is on the screen and parse complex images as prompts. That changes the input model from “type a precise request” to “show the assistant what you mean.”

A shopping list, inspiration image or visual reference can become the starting point for an action. The user no longer has to translate every visible detail into a sequence of text instructions before Gemini can begin.

Context-aware input is powerful because it reduces description work. It is also a reason to keep the user in the loop. Images can contain unrelated people, private messages, addresses or ambiguous visual details. A system that sees more needs to explain what it is using before acting on that context.

The strongest version of the experience is not “Gemini understands everything.” It is “Gemini shows enough of what it understood that the user can correct it early.”

How Gemini Notebook fits the foldable workflow

The same announcement brings Gemini Notebook, formerly NotebookLM, pre-installed on Samsung’s new foldables. Google presents it as a workspace for deep projects and source synthesis rather than quick task automation. Background: NotebookLM becomes Gemini Notebook, with code execution.

On the Galaxy Z Fold8 series’ larger display, users can drag photos, documents, whiteboard images and voice recordings into a side-by-side workspace. Gemini Notebook can then answer questions about those sources or create tailored resources such as slide decks, videos, flash cards, quizzes and podcasts.

That gives Samsung’s foldables two distinct AI roles:

• Delegation: Gemini handles a multi-step task across supported apps.
• Synthesis: Gemini Notebook helps the user understand and transform a collection of sources.

The distinction matters for product design. Delegation needs progress visibility, permissions and a final confirmation. Synthesis needs source grounding, editing controls and a way to inspect whether the output reflects the supplied material.

Google is also including a six-month Google AI Pro trial on the announced Galaxy Z Fold8 Ultra, Fold8 and Flip8 devices, with regional and eligibility conditions in the announcement.

How Galaxy Watch 9 extends Gemini access

Google says Gemini on the Galaxy Watch 9 can be activated by simply raising the wrist, without a wake word. The interaction is designed for moments when reaching for a phone is inconvenient: walking, carrying something, cooking or moving between tasks.

The wrist gesture is a small interface change with a larger implication. Wearables can become a low-friction trigger for an agent that completes work elsewhere. The watch does not need to display every step if the phone or notification system can show progress and request confirmation.

That also creates a new question for users: which device is responsible for explaining what the agent is doing? Cross-device assistants need consistent state. If a task begins on a watch, continues on a phone and finishes in a third-party app, the user should not have to reconstruct the workflow from separate notifications.

How Android XR glasses move Gemini beyond phones

The second headline from the event was physical rather than software-based. Google and Samsung shared more about upcoming Android XR intelligent eyewear and revealed two additional frame designs from Gentle Monster and Warby Parker.

Samsung’s Newsroom announcement describes the eyewear as a collaboration that combines Samsung’s hardware work with the partners’ design language. Google says the intelligent eyewear is being developed with Samsung and launching in the fall.

The collaboration addresses a practical problem for smart glasses: people must want to wear them before they can benefit from the intelligence inside them. A technically impressive device that looks like a lab prototype has a much smaller path into everyday life than a frame that fits a person’s style and routine.

Samsung describes a lightweight frame and slim profile, while the Google and Samsung announcements emphasize hands-free, context-aware assistance. The designs are therefore doing two jobs at once:

1. They make Android XR more visible as a consumer product category.
2. They test whether AI assistance can become part of ordinary eyewear rather than another screen people must hold.

What Gemini is designed to do on Android XR glasses

Samsung describes intelligent eyewear with a built-in camera for visual context and assistance that can respond to what the wearer sees. Google says Gemini will come to the glasses and that a compatible Wear OS watch can provide hand-gesture control.

Samsung’s announcement describes possible experiences including summarizing messages, live translation, finding a route, capturing a whiteboard and sharing a first-person view during a call. These are product claims and intended experiences, not evidence that every feature will be available in every market or at launch.

The design goal is clear: let the wearer ask for help while staying engaged with the physical world. The user could look at a sign, a route, a document or a conversation and request assistance without reaching for a phone.

The risk is equally clear. A camera and a context-aware assistant can be useful only if bystanders, privacy, recording and consent are handled as product requirements rather than afterthoughts. Hands-free does not mean consequence-free.

Why Gemini automation and Android XR eyewear reinforce each other

Phone automation and XR eyewear solve different parts of the same product problem.

The phone has the richer screen, more familiar controls and access to the apps where tasks actually happen. The glasses have a more natural relationship with the user’s surroundings and can reduce the need to look down at a screen.

Together, they suggest a distributed agent model:

The model is not the only product surface. The interface, permissions, notifications, context handoff and confirmation step determine whether the system feels helpful or unpredictable.

What more than 40 Gemini apps does not prove

More than 40 supported apps is a useful expansion, but it is not the same as more than 40 reliable workflows. A task can fail because a page changes, a login expires, a product is out of stock, an address is ambiguous or a booking requires a personal choice.

The better evaluation questions are:

• How often does Gemini complete the requested outcome correctly?
• How clearly does it show the next step before taking it?
• How easy is it to interrupt or correct the task?
• Does the user receive enough context to approve the final result?
• What happens when the supported app behaves unexpectedly?
• Can the user review which information was shared across apps?

The same principle applies to glasses. A camera-enabled assistant should be evaluated on whether it gives useful, accurate, socially acceptable help—not merely whether it can identify an object in a demo.

Google’s product announcement includes cautions about regional, device, language and age availability, as well as the need to supervise tasks and check responses. Those conditions are part of the product, not paperwork around it.

What Gemini task automation changes for Android developers

The announcement does not mean every Android app automatically becomes a Gemini automation target. It means supported app flows are becoming a strategic surface for delegation.

Developers who want their apps to participate in this ecosystem should think about the actions an agent can safely perform, the information it needs, the points where confirmation is required and how a user can recover from an incomplete task.

Good agent-ready app flows should make these things explicit:

• the user’s intended goal;
• the minimum data needed to perform the action;
• the state of the task;
• the reversible and irreversible steps;
• the final confirmation boundary;
• the error and recovery path.

An app that is easy for a human to navigate is not necessarily safe for an agent to operate. Agentic interfaces need clear state and predictable consequences.

What Google and Samsung are building toward

Google is turning Gemini Intelligence into a cross-device operating model. The first visible expansion is task automation across more than 40 apps on Samsung’s newest foldables. The next is a set of watches and Android XR glasses that let Gemini meet the user in more moments, with less dependence on a phone screen.

The strategic bet is that an assistant becomes more useful when it can do the work, see the context and appear on the device that fits the moment. But the quality bar rises with the convenience. Users need visible progress, easy interruption, clear permissions and a final confirmation step for consequential actions.

Samsung’s two new eyewear designs make the hardware side of that bet tangible. Gentle Monster and Warby Parker help Android XR look like something people might wear; Gemini is supposed to make it useful once they do.

The next phase of Android AI will therefore be judged less by how eloquently Gemini answers and more by whether it can move between apps, phones, watches and glasses without making the user lose control of the task.

FAQ

How many apps does Gemini task automation support now?

Google says Gemini task automation now supports more than 40 popular apps, expanding beyond the handful of rideshare and delivery apps available in the earlier beta. The announced categories include shopping, restaurant reservations, travel experiences and event tickets. That number describes supported app coverage, not a promise that Gemini can automate every app or every workflow. Google also notes that availability depends on supported apps, devices, markets, languages and age requirements, so the practical experience will vary by user. More on this: Gemini app plans after AI Studio cancellation.

Can Gemini task automation run in the background?

Yes. Google says users can close the phone or continue multitasking while Gemini works through a requested task. The user can track step-by-step progress, interrupt the process, adjust a specific detail and then hand control back. Gemini is designed to notify the user when the task is ready for review and final confirmation. That review boundary matters because navigating an app successfully is not the same as approving a purchase, reservation, message or other consequential action. See also Gemini Spark can now use Chrome to auto browse, AI Pro access goes international. Background: Gemini for macOS adds voice control and smart dictation. Background: Gemini Spark comes to AI Pro, not just Ultra. Background: Gemini Chrome for Android adds a paid task agent.

Which Samsung phones get these Gemini features?

Google highlighted the Galaxy Z Fold8 Ultra, Galaxy Z Fold8 and Galaxy Z Flip8 alongside the Gemini Intelligence rollout. On the Galaxy Z Flip8, users can open task automation from the Flex Window by long-pressing the power button. The announcement does not mean every feature has identical availability across all markets or devices. Google’s footnotes also say that supported apps, devices, languages and countries can affect access, so buyers should check the local product conditions before treating the announcement as a universal feature list.

What are the two new Android XR glasses designs?

Google and Samsung revealed two additional intelligent-eyewear frame designs from Gentle Monster and Warby Parker. The partnership is intended to combine Samsung’s hardware work with the eyewear brands’ styling, while Google positions the glasses as an Android XR device for Gemini assistance. The designs are upcoming rather than a currently available replacement for a phone. Google says the eyewear is launching this fall, but feature access and final availability can still depend on market, device and service conditions.

Will the Android XR glasses replace a phone?

Nothing in the announcements supports that conclusion. Google and Samsung describe the glasses as a hands-free extension of the Android and Galaxy ecosystem, with Gemini responding to visual context and a compatible Wear OS watch offering gesture control. Phones remain the richer surface for planning, inspecting, editing and confirming complex work. The more realistic near-term model is a distributed assistant: glasses supply context and quick interaction, while the phone remains the place where users can see more detail and make deliberate decisions.

What is the main risk with Gemini task automation?

The main risk is confusing successful navigation with successful completion. A user should be able to see what Gemini understood, which app it is using and what will happen next. They should also be able to interrupt the task, correct details and confirm the final action. Those controls matter most when automation touches money, reservations, messages, travel or private information. More supported apps can make Gemini useful, but only if the system keeps the user’s authority visible at every consequential boundary.