Google is testing a new feature in a trial basis. The new tool called ‘Call for me’ allows Gemini to call businesses for the user. The initial release of the update is limited to a few users. Only those Pixel 11 owners in the US who have subscribed to Gemini services can access the latest update. Additionally, users must have the beta version of Google’s Phone app for Android to use the feature. This restriction suggests that Google views the innovation as an experiment, not a finished product.
This development comes as other AI agents acquire similar capabilities. For example, Meta’s Muse and Instinct can call for users, and Google has implemented a comparable function in Gemini.
What Call for Me Actually Does
Google promises that the AI handles reasonably complex tasks. It can call a store and ask if something is in stock, book a restaurant, reschedule a visit to another day, or request that a business keep something on hold. During the call, Gemini announces itself, navigates through menus, waits on hold, and engages in conversation when an agent answers the call.
You are not required to have blind faith in the AI. There is an option for a live transcript so that you know exactly what is happening during the call, and you can always take over the conversation at any time. Additionally, the calls are made using your phone number, dialed from your phone, so the conversation does not appear to be coming from a Google number. That way, if the business calls back, it is more likely to reach you instead of a Google employee, although the announcement does not discuss call-back options.
The option to take over the call has to have some value. A conversation could veer off-topic, such as if an employee asks a question that the AI was not prepared for. Instead of ending the conversation and moving on to the next task, you could take over the call in the middle of the conversation. Because transcripts appear in real-time, you do not have to second-guess whether or not everything is going smoothly. You can simply read what is happening and decide for yourself if you need to do anything.
Four Tasks and Four Different Challenges
Google’s examples span a wide variety of tasks and have different requirements for the AI.
A simple check of whether a product is in stock is the easiest since it is a short question that only requires a short answer. The main challenge is to contact the right department. Making a restaurant reservation adds more complexity as the user wants to specify the party size, the desired time, and possibly the name of the booking. Rescheduling an appointment is a trickier task as one needs to identify an existing booking before finding a new fitting slot. Asking a store to hold items for a customer falls in the middle of the complexity spectrum: while it requires a positive response, the employee will have to take specific actions.
The video shows four tasks at different levels of complexity. Google aims to demonstrate the capabilities of its technology. However, the practical application and actual performance are what the demos are supposed to showcase.
Sharing Your Details Is the Big Change
Google has had AI assisted calling functions for some time now. However, previous incarnations of Google’s system only allowed for certain restrictions be placed on such calls. The new iteration does allow the sharing of personally approved information as part of the calling process. According to Google, that allows Gemini to take a broader range of actions.
Think about any sort of reservation or rescheduling function. It typically includes a name, time, phone number, and potentially one other detail that allows the receiving business to identify an existing booking. A calling agent that cannot pass any of this information on would find their utility extremely limited. This is the point where a calling assistant goes from asking questions to actually performing actions.
The TechCrunch article does not specifically mention what personally identifiable information Gemini would be able to relay, nor does it specifically state the process by which such information would be approved for sharing. Those two factors will dictate the level of confidence that a user would have when sharing any potentially sensitive information over the phone.
A Long Road From the Salon Demo
Google has been working on it for quite some time now. At their I/O developer conference in 2018, they unveiled a proof of concept that had Google Assistant calling a salon to make a reservation. They even added in a few ‘umm’s and ‘ahh’s to make the call sound more human. This ‘proof of concept’ was named ‘Call for Me’. It is presented as a tool to assist with everyday tasks.
An announcement last year revealed ‘Ask for Me’. This tool allows Gemini to call businesses and ask for things like pricing and services. Pixel owners have had other calling tools at their disposal as well. ‘Hold for Me’ and ‘Talk to a Live Rep’ can work through menus and hold until a live person answers the call. ‘Direct My Call’ takes a less involved approach, showing which numbers to press when menus are too difficult to decipher.
In an ordered list, it’s easy to see how each tool progressed from the last. The early tools took on the busy work, such as menus and waiting. ‘Ask for Me’ added simple questions. ‘Call for Me’ builds upon that, allowing Gemini to take control of the conversation with your approved details.
Why the Rollout Is So Narrow
The eligibility list is short. You need a Pixel 11, a US location, a paid Gemini subscription, and a beta Phone app. Google cited “the complexities of real-world conversations” as a reason why this experiment is happening on a small scale, and it will likely take some time for this feature to roll out more widely.
That’s not surprising given the potential edge cases of a conversation on the phone. Someone could ask for something odd, talk fast, or get sent to another person. A smaller group of people could give Google enough data from live calls to figure out how to best let its AI assistant handle the chaos of everyday conversation.
The fact that a subscription to Gemini is required cuts down on the possible pool of testers. Google has not stated if the feature will remain behind a subscription after the experiment wraps up.
Another reason this approach makes sense is that it allows Google to maintain control of the experiment. The feature is built into the Google phone app and uses Google devices to make calls. There’s less room for error than there would be if the feature was built into a third-party phone or another app.
What Google Hasn’t Said Yet
As mentioned, the announcement leaves many practical questions unanswered. It does not say how much the Gemini subscription would cost for this purpose or when other phones and regions will be able to use this feature. Nor does it explain how a business will know it is talking to an AI or if transcripts of the conversation are saved once the call ends.
Some of these details may be as important to the businesses that receive calls as the ones who make them. A restaurant host who answers a call from an AI to make a reservation will have his or her opinions on how it should act. So far, nothing in Google’s announcement touches on that aspect of the call.
The Bigger Picture for AI Agents
Calling was a showpiece demo once. It’s becoming a default entry for personal AI agents’ feature lists. Muse, for instance, is Meta’s personal agent that makes links between a user’s apps and services, such as email, and calendars, while phone calls are a natural part of that kind of assistant.
Google is different when it comes to its controls. Live transcripts, a takeover option, and approval over shared information all keep the user in the loop. If people actually listen to the transcript or just sit and burn the call is another matter that Google will play to learn as the trial continues.
For now, Call for Me is a small-scale test for a limited group of Pixel owners Google has the history, the device, and the phone application to implement it. What it will still lack is confirmation that actual conversations will trend as their simulated counterparts do.
FAQs
Q1: What is Call for Me?
It is a new Google experiment that lets Gemini call businesses for you. It can ask about stock, book a meal, change appointments and more.
Q2: Who can use it?
Pixel 11 owners in the US who are paying for a Gemini subscription. In addition, they must have the beta version of Google’s Phone app for Android.
Q3: Can I listen in on the call?
Yes. You can view a live transcript of the call as it happens. You can also take control of the call at any time.
Q4: Does Gemini use my phone number?
Yes. Google states that Call for Me uses your number, and that the call is initiated from your phone.
Q5: Can Gemini share my personal information?
Yes. However, you can approve what information can be shared. Google says that this allows Gemini to take a much wider range of actions than previously possible with calling features.
Q6: How is this different to Ask for Me?
Ask for Me, a similar Gemini experiment that launched last year, could call businesses to ask for things such as prices and availability. Call for Me can perform more complex tasks and share approved personal information.
Q7: Can Gemini deal with hold music and phone menus?
Yes. Gemini can navigate through phone menus and wait on hold until a real person answers.
Q8: Why is Google rolling this out gradually?
Google says that conversations can be “nuanced”. It wants to release the feature more widely after testing it.
Q9: Are there other AI assistants that can make calls?
Yes. Meta’s Muse and Instinct can make calls on behalf of the user, according to Tech Crunch. Google’s new feature launches as these assistants get the same ability.
Q10: Can I use Gemini’s calling feature on another phone?
Google hasn’t confirmed that. For now, the feature is only available to Pixel 11 owners in the US on a paid Gemini plan. Google says that it wants to spend time perfecting the feature before rolling it out more widely.
