September 24, 2026
TechCrunch reported this week that Google is testing a feature letting Gemini call businesses on your behalf, starting with Pixel 11 owners in the U.S. who pay for a Gemini subscription. Picture asking your phone to check a restaurant's hours or reschedule an appointment, and having an AI voice place that call while you do something else. It's a small rollout -- one phone model, one country, one paying tier -- but it's a meaningful signal about where Google thinks voice AI is headed next: not just answering your questions, but acting as your proxy in the outside world.
For businesses, this is the part that should get attention. If a growing share of inbound calls are actually AI agents calling on behalf of customers, the humans (or AI systems) answering those calls need a way to know that quickly, and to route them appropriately. A restaurant host or a scheduling desk built around human-to-human conversation patterns is going to get confused fast. Any business running phone-based intake, whether that's a helpdesk or a CRM-driven sales line, should start thinking now about how their workflows handle a caller that's a bot rather than a person.
The same day, TechCrunch published a conversation with the CEO of ElevenLabs, the voice-AI company reportedly now valued at $22 billion, whose technology already powers a lot of the synthetic voices on the other end of customer service calls. The CEO's own framing was telling: businesses probably should tell customers they're talking to an AI, he said -- at least until that becomes the default assumption anyway. That's a remarkably candid admission from someone whose company profits from the ambiguity. It's also, in my view, backwards. "Should" disclose is doing a lot of work in that sentence, and it's the same hedge every industry makes right before regulation forces the issue anyway.
Put Gemini's calling feature next to ElevenLabs' valuation and you get a clearer picture of where voice AI actually sits right now: it's good enough, cheap enough, and normalized enough that both ends of a phone call could plausibly be synthetic within a couple of years. That's not inherently bad -- AI handling routine calls can save real time for both sides. But it raises a fairness question that neither company has fully answered: if your AI is calling my AI, who's actually accountable when something gets promised, missed, or misunderstood? Voicemail transcription errors are annoying. A missed appointment or a garbled price quote booked by an AI acting as an unaccountable proxy is a real cost. Businesses building AI into customer-facing workflows, whether through marketing automation or support lines, would do well to keep a clear disclosure and audit trail baked in rather than treating transparency as optional polish.
Meta's Muse Charm and Google Photos' Clueless-style virtual closet, both also reported by TechCrunch this week, look unrelated to enterprise voice AI on the surface, but they share a thread: AI is quietly becoming ambient rather than something you consciously open an app for. A wearable gadget shaped like a nostalgia toy and a wardrobe built automatically from your camera roll are both bets that AI works best when you don't have to ask for it. That's the same bet Google is making with Gemini calls -- AI acting on your behalf, in the background, with minimal friction. The pattern is consistent even when the products are playful. The businesses that get ahead of this won't be the ones chasing every gadget, but the ones that quietly automate the repetitive parts of their own operations first.
Where do you land: would you want to know if the person calling your business is actually an AI, or does it only matter if the call goes wrong?
Sources