September 25, 2026
TechCrunch reported that Google is testing a feature letting Gemini call businesses on a user's behalf, starting with Pixel 11 owners in the U.S. who pay for a Gemini subscription. On the surface this reads like a novelty -- your phone makes the awkward call to the dentist so you don't have to. But the real story is what it implies about trust and liability once an AI agent starts transacting with the outside world instead of just answering questions inside an app. We wrote about this the day it broke, and the questions haven't gotten easier: if Gemini books the wrong appointment, cancels the wrong reservation, or misstates your request to a human on the other end of the line, who's accountable -- Google, the user, or the business that answered? For now this is a narrow, opt-in Pixel feature, so the stakes are low. But it's a preview of where every major assistant is headed: from answering to acting. Any business owner using AI tools internally should watch this closely, because the same accountability question shows up the moment you let an AI-built internal tool place orders, send invoices, or contact customers without a human double-checking first.
Two stories today point at the same shift from a different angle. TechCrunch described Meta's new Muse Charm as a dangling, Tamagotchi-shaped AI gadget that's really tapping into a Gen Z trend of turning wearable tech into fashion accessories. Separately, PrismML announced it's bringing its small, open-weight language models to Qualcomm-powered smart glasses, betting that on-device AI can make better use of hardware that's already sitting on someone's face. These are different bets on the same thesis: the next AI interface isn't a browser tab, it's something you wear. Meta is playing the consumer-culture angle -- make the AI gadget desirable the way a phone case or a keychain is desirable. PrismML is playing the infrastructure angle -- make small models efficient enough to run locally, without shipping every request to a cloud server. Both are legitimate strategies, but they're built for very different buyers. A fashion-forward gadget is a hard sell into a business budget. Efficient, on-device models that cut cloud costs are a much easier one, and it echoes a trend we've flagged before around cheaper, leaner AI hardware. For companies evaluating where to put their own AI investment, the PrismML approach is the one worth studying: smaller, cheaper, local models solving a specific job well, rather than a flashy device looking for a use case.
Almost lost in today's headlines is Google's announcement that its Clueless-inspired virtual closet feature -- which builds a wardrobe from your photo library -- is now broadly available on Android and iOS after piloting on Android since June. It's a fun consumer feature, but it's also a clean example of a pattern worth stealing: Google didn't build a new app, it found an existing pile of data (your camera roll) and layered a narrow, useful AI feature on top of it. That's the same logic behind good internal business tooling -- the data you already have (customer records, project history, support tickets) usually holds more value than another standalone app would. It's the same argument we'd make to any team stuck choosing between bolting on another point solution and building something that actually uses the data they've already got, whether that's through a CRM or a purpose-built internal tool.
TechCrunch also reported that Lightspeed is targeting $250 million for a new India-focused fund aimed at early-stage AI, and for the first time aligning its India fundraising cycle with its global funds while shortening its investment period. That's a signal worth noting alongside the wearable and agent stories above: venture capital isn't just chasing frontier-model labs anymore, it's chasing the application layer in markets outside the usual U.S.-China axis. A shorter investment period suggests Lightspeed expects these AI bets to prove themselves -- or fail -- faster than a typical venture timeline. For founders, that's a tighter window to show traction. For everyone else, it's a reminder that the AI wearable and agent race described above is just the visible tip of a much bigger, faster-moving capital cycle underneath it.
If Gemini starts making calls on your behalf, would you actually trust it to handle something that matters -- or only the errands you'd rather skip?
Sources