The restaurant has a table, but it wants a card number to hold it. Your agent was asked to make a reservation. What happens next? That ordinary turn in a phone call is where I would look for an advantage in the contest between Google and Meta. My forecast is that Google’s visible review and takeover controls could earn more repeat permission to call on a user’s behalf. The companies still have to demonstrate how their designs behave in that moment.
TheStreet’s September 27 comparison is the starting point for this opinion about Google’s limited phone beta and Meta’s Muse test.
Google offers controls at the moment of the call
Google’s new Gemini Call for Me trial is narrow. It is gradually reaching adults in the United States who have a Pixel 11, a Google AI subscription, and the public beta of the Phone app. The agent uses the user’s phone number and mobile network. Google says it identifies itself as an AI assistant on a recorded line, shows the task and information to be shared before dialing, permits live listening or transcript monitoring, and lets the user take over. It also rules out payments and sharing sensitive financial or health information.
Review before dialing, observation during the call and takeover give a person three chances to intervene. TheStreet’s comparison puts Google beside Meta’s Muse. My interest is in whether those controls remain easy to find while the conversation is moving. A button that lets you correct the reservation before it is agreed can be more useful than a fluent explanation afterward.
The 72-eggs story puts the same boundary in a grocery cart, where an unexpected extra can survive until checkout if the person only reviews the total.
A human handoff needs its own disclosure
Meta’s Muse also promises to act across a person’s services, and its launch materials describe a protected virtual machine, user controls, and an audit trail. Those features deserve evaluation. But earlier Newsroom reporting on Muse and Meta’s privacy record documented a problem close to this use case. Internal testers encountered calls described as AI work that were completed by human callers without the human handoff being clear to at least one tester. Meta said the feature was still being tested and that a public launch would include appropriate disclosure. The share of calls routed to people remains unknown.
The earlier article also examined Messenger audio sent to contractors and Meta’s stated use of some AI conversations for advertising. Those histories give people a reason to ask exactly who will receive a call request and its personal context. A name, an appointment time or a preference may be ordinary on its own, yet the user should still be able to choose whether it reaches a human operator. A visible handoff would make that choice possible.
The transcript has a life after the call
Google’s own data practices deserve the same attention. Its Phone app can hold a transcript and recording, so retention, review, deletion and secondary use belong in the decision to try it. Google’s Search agentic calling supplies experience with local-business inquiries. Performance in that setting does not measure this Pixel beta against Muse. The hardware and subscription limits also give the trial a relatively narrow starting audience.
Google’s existing phone surface and explicit intervention controls give it a plausible opening. Meta’s reach gives it opportunities to improve a calling experience in front of many users. I expect the product that makes delegation easiest to inspect and correct to earn the next call. That is a forecast about repeat permission in agentic AI , with errors and actual use still capable of changing the outcome.
What happens when the restaurant asks another question
When a chatbot suggests a reservation time, you can reject it before anyone else hears it. A calling agent is already speaking to the person who can accept the booking. A summary after the call may arrive after the important choice. The approval screen and takeover control put the user back into that sequence at points where a decision can still change.
Return to the restaurant’s request for a card number. Google’s published limits exclude payments and sharing sensitive financial information. The useful behavior would be to stop or return control when the conversation reaches that limit. A real call may wander beyond the original request, so the system needs to recognize the boundary while still explaining what the user can do next.
Meta can make the human handoff just as concrete. Before sending any information, show whether a model or a person will place the call, give the user a choice and keep a usable record. A protected virtual machine and audit trail address other parts of the system. The person delegating the errand also needs to understand who is going to speak in their name.
Follow the unplanned request through the call
The restaurant example gives a proposed evaluation a small, observable sequence:
- Compare the approved booking task with the new request for financial information.
- Observe whether the system stops, explains the limit or returns control.
- Check whether the resulting record identifies who spoke and which decision the user made.
This is a proposed comparison, not a claim that Newsroom ran the beta or measured either provider's error rate. It makes the trust forecast open to evidence.
What would change my forecast
Four observations would move my judgment. Do businesses consistently hear who is calling? Can users review, monitor and interrupt the task? Does the record make errors, cancellations and human handoffs clear? Can a person understand what happens to the recording and transcript afterward? Public evaluations and ordinary use over time can answer those questions in a way launch attention cannot.
The next invitation to delegate a call will be a small decision: was the last one easy to inspect, and did I stay in control when it changed? Google could lose that permission through repeated errors or unclear data use. Meta could earn it with a clearer handoff. The takeover button is one place where those competing promises become an experience someone can judge.
FAQ
What should a booking agent do if the restaurant asks for my card details?
If sharing that information exceeds the permission you gave, it should bring the request back to you. Review the exact information and take over if necessary.

