1. Headline & intro
Gemini’s latest trick on Android looks modest on paper – ordering food or a ride for you – but strategically it’s a big step toward something more radical: a phone that does things for you instead of waiting for you to tap through apps. If Google pulls this off, Android stops being a grid of icons and becomes an operating system for AI agents. In this piece, we’ll unpack what Google actually shipped, why it matters far beyond food delivery, how it reshapes the Google–Apple–OpenAI race, and what it could mean for users and regulators in Europe.
2. The news in brief
According to TechCrunch, Google has introduced new Gemini-powered automations on Android that can complete certain multi-step tasks, such as ordering an Uber or arranging food and grocery delivery. The feature is currently in beta and only works with a limited set of partner apps in those categories.
Support is restricted to the Gemini app on specific high-end phones: Google’s Pixel 10 and 10 Pro, and Samsung’s Galaxy S26 series. Availability is initially limited to users in the U.S. and South Korea.
Alongside automations, Google is expanding Gemini-based scam detection for calls and texts. Call scam detection, previously on Pixel devices in several countries, is rolling out to the Galaxy S26 in the U.S., while on-device models will also analyse texts for scam patterns in selected markets.
Finally, Google’s Circle to Search feature is being upgraded so it can identify and search for multiple items visible on the screen at once, not just a single object.
3. Why this matters
Multi-step automation is where AI stops being a chat toy and becomes infrastructure. Until now, consumer AI assistants have mostly produced text, images or code. Gemini on Android is starting to produce actions in the real world: rides booked, meals ordered, tasks completed without you opening half a dozen apps.
The winners are clear. Google gains a powerful lock‑in mechanism: if Gemini is the layer that actually “gets things done” on your phone, switching to a rival platform becomes painful. App developers in supported verticals may also benefit from higher conversion, as AI funnels users directly into completed orders instead of abandoned carts.
But some players lose control. Consumer brands that have invested heavily in app design suddenly find that an AI agent is front‑ending their service. The user doesn’t care if the ride came from Uber or Lyft; they care that “Gemini got me a ride fast.” That commoditises the underlying services and strengthens the platform owner – classic Google.
For users, the trade‑off is convenience versus transparency and reliability. A visible progress UI and the ability to stop a task, as described by TechCrunch, are important safeguards, but they don’t remove the core risk: an AI that can click and swipe on your behalf is only as trustworthy as its training, guardrails and sandboxing.
And in the background, this escalates the competitive pressure on Apple, which is already struggling to modernise Siri. If Android becomes the “automation OS” first, Apple’s reputation for a smoother UX may not be enough to compensate.
4. The bigger picture
Gemini’s automations land in the middle of a broader shift toward AI agents that can operate computers, not just chat. OpenAI has been experimenting with scheduled and scripted tasks in ChatGPT, Anthropic is positioning Claude as a co‑worker that can manipulate files and workflows, and smaller tools like OpenClaw (mentioned in the TechCrunch piece) have gone viral by handling email, calendars and travel logistics.
Historically, we’ve seen consumer automation waves before: think IFTTT, Tasker on Android, or Google Assistant routines. All of them hit the same wall: brittle rules and narrow integrations that normal users couldn’t be bothered to maintain. Large language models change that equation. They can interpret messy, natural instructions like “reorder what I bought last week but skip the snacks” and figure out the necessary sequence of taps.
What Google is doing now is essentially turning that LLM intelligence into an OS primitive. The “secure virtual window” that runs automations is conceptually similar to a headless user session. If this architecture proves safe, it scales well beyond food and rides: banking, travel, health, productivity.
Competitively, this pushes Android toward the same frontier that Windows and the web will face: once AI can reliably drive the UI, the distinction between “mobile app”, “website” and “service” starts to blur. Google, with its control of the OS and the Play ecosystem, is trying to ensure that blurring happens in its favour rather than, say, in favour of a cross‑platform OpenAI agent.
5. The European angle
For European users, the first point is simple: this is not launching here yet. The beta is U.S.‑ and Korea‑only, and even scam detection’s expanded rollout is U.S.‑first on Samsung’s S26. But that doesn’t make it irrelevant for Europe; it makes the continent the test case for how AI automation can coexist with stringent regulation.
Gemini’s ability to operate apps on your behalf raises immediate GDPR questions: what data is processed on‑device versus in the cloud, how consent is obtained for scanning screens and messages, and how users can exercise rights like access, correction and deletion over automation logs.
The EU’s Digital Services Act and upcoming AI Act add further layers. An AI that places orders or interacts with third‑party services could be seen as making “meaningful decisions” that affect consumers financially. That invites transparency obligations: clear labelling that an AI, not a human, executed the action; easy ways to contest mistakes; auditability for regulators.
Then there is the Digital Markets Act. Google is a designated gatekeeper, and an OS‑level agent that preferentially routes tasks through certain apps or services will draw scrutiny. If Gemini routinely favours Google’s own services or select partners, expect European competition authorities to ask whether rivals get fair access to the automation layer.
For European startups building vertical assistants – in fintech, travel, health – this is both threat and opportunity. The threat: Google may own the primary interface. The opportunity: if Google provides robust, non‑discriminatory APIs for agents, European players can plug in as specialised backends rather than trying to own the entire user experience themselves.
6. Looking ahead
In the near term, expect three things. First, rapid expansion of supported categories: travel bookings, recurring subscriptions, maybe even simple admin like changing a flight or filing a return. Food and rides are just the low‑risk sandbox where errors are annoying but rarely catastrophic.
Second, a battle over automation standards. If every platform builds its own private protocol for AI agents to drive apps, developers will be stuck implementing five different “AI control” SDKs. The more interesting scenario is some form of open or semi‑open schema – think of a future where apps declare machine‑readable flows (“order”, “cancel”, “reschedule”) and agents from Google, Apple, or others can safely orchestrate them.
Third, friction with regulators will increase as soon as money moves. What happens when an AI agent orders from the wrong merchant, or a scammer tricks Gemini into confirming a fraudulent transaction? Liability will be contested between the platform, the app and possibly the user who “approved” the automation. Insurers will quickly get involved.
On a 2–3 year horizon, the larger risk for Google is user trust. One spectacular automation failure – a viral story about a wildly wrong order or a security mishap – could slow adoption. The opportunity, however, is enormous: if most people come to rely on an Android agent for daily logistics, Google will have created the stickiest service it has launched since Gmail, deeply woven into routines and hard to abandon.
For European readers, the key milestone to watch is the first official announcement of Gemini automations for EU markets. When that happens, look closely at the fine print: where processing happens, which safeguards are EU‑specific, and whether Google positions this as an open ecosystem for third parties or a tightly curated club.
7. The bottom line
Gemini’s new Android automations are limited in scope today, but strategically they mark the shift from AI as a conversation partner to AI as an operator of your digital life. If Google can scale this safely and transparently – and convince European regulators it plays by the rules – Android could become the first mainstream “agent OS”. The open question for readers is simple: how much of your daily decision‑making are you really prepared to outsource to an invisible assistant that taps the screen on your behalf?
