Calling a business to check whether an item is available, change a reservation or move an appointment can take only a few minutes but the waiting, automated menus and back-and-forth can make the task feel far more time-consuming.

Google wants Gemini to handle that work instead.

The company is testing Call for Me, an experimental Gemini-powered calling feature that allows the AI assistant to place certain calls on a user's behalf and participate in the conversation.

Rather than simply giving users instructions on what to say, Gemini can actually make the call using the user's phone number, communicate with the business and return the relevant information to the user.

The feature represents another step in Google's broader effort to turn Gemini from an AI assistant that answers questions into one that can complete tasks in the physical world.

Gemini Doesn't Just Tell You What to Say

The difference between Call for Me and a conventional voice assistant is what happens after the phone number is dialled.

With the new experience, a user could ask Gemini to call a hardware store and find out whether a particular product is available.

The assistant could also be used for tasks such as changing or cancelling a reservation or adjusting a scheduled appointment.

Once the user authorises the task, Gemini handles the phone conversation rather than simply preparing a script for the person to read.

That distinction is important.

AI assistants have been able to generate emails, draft messages and provide suggested responses for years. Google's latest approach is about allowing the assistant to act on the user's behalf.

Users Can Watch the Conversation

Giving an AI permission to speak to businesses creates an obvious trust problem: what exactly is the assistant saying?

Google's solution is to make the interaction visible.

During the call, users can see a transcript of the conversation in real time.

They can also take over the call when necessary.

That gives the user a way to intervene if the conversation moves in an unexpected direction or if the business asks a question Gemini cannot handle.

The approach effectively puts a human in the loop rather than treating the AI as an entirely independent phone operator.

Google Is Keeping the Experiment Narrow

Despite the potential, Google is not presenting Call for Me as a fully autonomous telephone agent that can handle any conversation.

The company describes the experience as an experiment and has placed restrictions around what it can do.

One limitation is that Gemini may struggle when a conversation moves outside the expected scope of the original task.

For example, if a user provides a highly detailed request and the business representative asks an unexpected follow-up question, Gemini may not be able to respond appropriately.

That limitation highlights one of the biggest challenges facing agentic AI.

Understanding a written request is one thing.

Handling an unpredictable conversation with another human is another.

Gemini Can Suggest What to Say

Google is also giving users additional control by allowing them to provide example phrases or instructions that Gemini can use during the call.

This can help the assistant understand how the user wants the interaction to proceed.

It also reflects Google's attempt to balance automation with user control.

Instead of simply saying, "Call this business," users can give Gemini more context about what they want accomplished.

The more specific the task, the easier it becomes for the system to stay within the boundaries of the conversation.

The Feature Builds on Google's Long Phone-AI History

Google has been working on AI-assisted phone calls for years.

One of its most notable projects was Duplex, which demonstrated how machine learning could make phone calls to businesses to complete tasks such as restaurant reservations.

Google later brought related capabilities to Pixel phones through products such as Hold for Me and Direct My Call.

Hold for Me was designed to recognise when a user was placed on hold and notify them when a representative returned.

Direct My Call, meanwhile, could transcribe automated phone menus and display the available options on screen.

Google has also continued developing AI-powered calling and business-search tools. Its agentic calling feature in Search and AI Mode can already call local businesses to check product availability and gather information for users.

Call for Me takes the concept another step forward by putting Gemini more directly into the conversation.

Why This Matters for AI Assistants

The evolution is significant because it changes what users expect from an AI assistant.

Traditional assistants largely operated as information tools.

You asked a question.

They gave you an answer.

Modern AI agents are increasingly being designed around a different model:

Tell the AI what you want done, and let it figure out the steps.

Calling a business is a particularly useful demonstration of that concept because the task involves something AI cannot simply complete through a webpage or app.

A store may only provide inventory information over the phone.

A restaurant may require a phone call to modify a reservation.

A small business may not have an online booking system at all.

In those situations, the telephone remains the interface.

Google is effectively trying to make Gemini the person operating that interface.

But Human Conversations Are Still Difficult for AI

The feature also exposes a major limitation of current AI agents.

Real conversations are messy.

People interrupt one another.

They ask unexpected questions.

They change their minds.

They misunderstand information.

A business representative may also require details that were not included in the original instruction.

An AI agent that works perfectly when a conversation follows a predictable script can quickly become unreliable when the interaction moves outside that script.

Google's decision to restrict Call for Me during its early testing therefore makes sense from an engineering perspective.

The company can learn how people use the feature, where conversations break down and when users need to take control.

Access Is Currently Highly Limited

Call for Me is not currently a general Gemini feature available to everyone.

The early test is limited to selected users in the United States.

The requirements described for the experiment include a Google Pixel 11, a US SIM card, a US phone number, a qualifying Google AI subscription and participation in the Phone by Google public beta.

Google's Pixel 11 family launched in August 2026, with the company's latest phones designed around what Google calls Gemini Intelligence.

That means most Gemini users — including those outside the United States — will not currently be able to use Call for Me.

The limited rollout also gives Google room to test the technology before considering broader availability.

The Bigger Competition Is About Action

The AI assistant race is increasingly moving beyond who can produce the best written answer.

The bigger question is becoming:

What can the assistant actually do for you?

Google already has access to several layers of the technology required to answer that question.

It controls Android.

It operates Google Search and Maps.

It has a massive business directory through Google Business Profile.

It owns the Gemini AI platform.

And with Pixel, it controls both the hardware and software experience.

That combination gives Google an opportunity to connect AI reasoning with real-world actions in ways that standalone chatbots may find harder to replicate.

Call for Me is one example of that strategy.

From "Ask Gemini" to "Do It for Me"

The most important change may therefore be the language people use when interacting with AI.

Instead of:

"What should I say when I call the store?"

The future assistant could be:

"Call the store and find out if they have it."

Instead of:

"How do I change my reservation?"

It becomes:

"Call them and move my reservation to Saturday."

That is a much more ambitious role for an AI assistant.

It also requires considerably more trust.

The assistant has to understand the user's intent, communicate accurately, know when it can make decisions and recognise when it should stop and hand the conversation back to the human.

Google's current experiment is an early attempt at that model.

Call for Me shows where Google's Gemini strategy appears to be heading: from an AI that talks to an AI that acts.

The technology is still limited, and the current rollout is tightly controlled. That's important because phone conversations are unpredictable, and mistakes could have real-world consequences.

But the concept itself is powerful.

A huge amount of everyday administrative work still happens through phone calls. If AI agents can reliably handle those interactions, users could eventually delegate everything from appointment changes to product enquiries without spending time listening to hold music or navigating automated menus.

Google has been building toward this for years through Duplex, Hold for Me and Direct My Call.

With Gemini, those individual capabilities are beginning to look less like separate phone tricks and more like pieces of a larger strategy: make the AI assistant capable of getting things done on your behalf.