October 17, 2026
Speaking at TechCrunch Disrupt, Amazon devices chief Panos Panay laid out a future where you stop opening apps altogether -- ambient AI devices sit in the background, anticipate what you need, and act before you ask. It's a seductive pitch, and it's not a new one; Humane, Rabbit, and a dozen other 'AI companion' hardware bets have already tried and mostly failed to deliver it. What's different now is who's saying it. Amazon isn't a startup burning venture money on a novelty gadget -- it's a company that already owns the device layer in millions of homes through Alexa and Echo. If anyone can make ambient AI feel inevitable rather than gimmicky, it's a company with that kind of distribution. But here's the catch for any business reader nodding along: apps aren't going away because a consumer gadget gets smarter. Your CRM, your billing system, your support queue -- those are built on structured data and workflows that someone still has to configure, audit, and own. Ambient AI might change how a customer discovers your business or reorders from you, but it doesn't replace the backend doing the work. If anything, the more invisible the front end gets, the more your internal systems need to be solid, because there's no UI left for a human to notice when something's broken. That's as much an argument for owning a dependable internal tools layer as it is for betting on the next ambient gadget.
On the opposite end of the hype spectrum, Amulya Vishwanath -- who spent seven years running product marketing for NVIDIA's robotics platforms before moving into venture investing -- told TechCrunch that robotics is nowhere near its 'ChatGPT moment.' Her framing is blunt: we're in a 'Lego era,' where even the basic pieces -- actuators, robotic hands, safety standards, the software stack -- are still being rebuilt from scratch. That directly contradicts the narrative, popular among some foundation-model boosters, that one sufficiently general AI model is about to make robots competent at everything, the way large language models suddenly made text generation usable overnight. I think Vishwanath has the better argument, and not just because she's closer to the hardware. Language models could leap forward because text, code, and images are digital and infinitely copyable -- a breakthrough architecture can be retrained and redeployed everywhere instantly. Physical actuators wear out, safety certifications take years, and a robotic hand that works in a lab demo still has to survive a warehouse floor. That's a supply chain and manufacturing problem as much as an intelligence problem, and those don't compress on the same timeline as a model release. For businesses evaluating automation investments, the practical takeaway is to treat near-term robotics pitches with real skepticism, and to look instead at software automation -- the kind that actually is having its ChatGPT moment -- for near-term returns, something workflow automation tools are already delivering today without waiting on a robot that can tie its own shoes.
Put these stories side by side and you get a useful gut check for anyone making AI purchasing decisions right now. Panay is describing a future that sounds close because the software underneath it -- language models, prediction, personalization -- genuinely is advancing fast. Vishwanath is describing a future that sounds close in press coverage but isn't, because the physical bottlenecks haven't moved nearly as fast as the models have. The honest version of 2026 AI strategy is picking which camp a given pitch actually belongs to before you write the check. Software-layer AI, including the kind that runs your sales pipeline or your support desk, is worth adopting now; anything promising to replace physical labor on a short timeline deserves more patience. I'd rather see a company spend this year automating the workflows it already runs through a CRM than waiting for either a mind-reading gadget or a general-purpose robot to arrive and do it for them.
Which pitch do you find yourself more skeptical of right now -- ambient AI devices that claim to replace your apps, or robotics companies claiming a breakthrough is just around the corner?
Sources