Google has been building toward this for years. Back at Google I/O, the company famously demoed Google Assistant calling a hair salon, complete with convincing “umm” and “ahh” sounds to pass as human. That demo got a standing ovation and then mostly sat on a shelf. Now, according to TechCrunch, Google is testing something far more capable: a feature called “Call for Me” that lets Gemini make phone calls to businesses on your behalf, handle the full conversation, and report back.
The feature is launching in a limited experiment for Pixel 11 owners in the U.S. who subscribe to Gemini Advanced. It also requires the beta version of Google’s Phone app for Android, which signals that Google is treating this as a controlled test, not a wide release. That’s a reasonable call. Phone calls are unstructured, unpredictable, and often frustrating. Getting an AI to handle them gracefully is a harder problem than it looks.
What makes Call for Me meaningfully different from earlier efforts is scope. Previous tools like “Hold for Me” or “Talk to a Live Rep” handled the mechanical parts of calling: navigating phone trees, waiting on hold, signaling when a human picked up. “Ask for Me” went a step further and let Gemini call businesses to ask basic questions about pricing or availability. Call for Me is supposed to handle the full interaction, including tasks like:
- Checking whether a product is in stock at a specific store
- Making or rescheduling a restaurant reservation
- Placing items on hold
- Navigating automated menus and waiting on hold, then continuing the conversation with a real person
Users stay in the loop throughout. The call happens from your actual phone number, dialed directly from your device, and a live transcript appears on screen so you can follow along. You can also take over at any point if the conversation goes somewhere unexpected. That’s an important design choice. Full autonomy with no override would be a hard sell for most users, and it would raise real questions about accountability when things go wrong.
The privacy angle is also worth noting. Google says Gemini can share personal information during calls, but only information you’ve explicitly approved. That’s vague enough to warrant scrutiny when more details emerge about what that approval process actually looks like in practice.
For context, this fits into a broader push from AI companies to build agents that take real-world actions rather than just answering questions. Meta has been working on similar territory with its Muse and Instinct agents. Apple has been slow here despite years of Siri promises. OpenAI has been moving toward agentic capabilities through its Operator product. Google’s advantage is that it already owns the phone app, the device relationship with Pixel hardware, and years of accumulated calling infrastructure. That’s a reasonable head start.
But the real question isn’t whether the demo works. It’s whether it holds up when a store rep goes off-script, when an automated system doesn’t behave as expected, or when the call requires judgment calls the AI wasn’t trained for. Google’s own framing is telling here. The company says it’s starting small because “real-world conversations are nuanced.” That’s an honest admission, and it’s the right instinct. The technology is interesting. Whether it’s reliable enough to actually save people time is a question that only a real rollout will answer.
