Summary: Amazon has launched two AI wearables under Bee, the startup it acquired in 2025: the wrist-worn Frontier ($74.99) and the card-sized, phone-mountable Eclipse ($99.99). Both listen to your daily conversations, then turn them into summaries, transcripts, and to-dos — and, if you opt in, surface product suggestions on Amazon via Alexa+ integration. For builders, this is a live case study in always-listening consumer hardware: the retention hooks, the monetization split, and the privacy tradeoffs are all on display.
What the devices actually do
After activation, Bee listens to conversations, combines what it hears with connected app data, and produces summaries, transcripts, insights, and action items in a phone or desktop app. Basic features include conversation summaries and a 14-day memory window with no subscription required. A $11.99/month Bee Pro tier adds unlimited recent-conversation summaries, suggested to-dos, daily memories, a personal chat assistant, templates, third-party integrations, and a desktop app. Per The Verge, the headline Pro feature is Amazon shopping and Alexa+ integration.
The two devices split by use case. Frontier wears like a Fitbit or clips to clothing, has dual microphones, USB-C charging, IP67 water resistance, and roughly 7 days of battery. Eclipse is not a wearable — it magnets onto the back of a phone, targets more professional contexts, records calls (with the other party’s consent required), and lasts over 21 days per charge, according to Android Central.
The commerce hook is opt-in — and that’s the interesting design choice
Shopping integration requires explicit user activation. Once enabled, Bee picks up products and preferences mentioned in everyday conversation — your partner asking you to buy cat litter, a friend recommending a book — and surfaces them as reminder cards on the Amazon homepage and inside Bee’s apps.
That’s a deliberate structure: the hardware is cheap (under $100), basic memory is free, and the subscription’s main draw routes you into Amazon’s storefront. It’s ambient commerce where the capture layer is a microphone on your wrist. If you’re designing an agent that acts on passive context, note how Amazon separated the sensing (default), the intelligence (free tier), and the monetizable action (opt-in Pro).
Privacy: good controls, unresolved third-party problem
Amazon’s controls are real: one-tap on/off with an LED indicator, auto-stop after prolonged silence, no stored audio (speech is converted to reviewable text and deleted), and transcripts encrypted with keys only the user’s device controls. Bee learns only the owner’s voice unless others consent. But SiliconANGLE points out that all data still uploads to Amazon, and privacy advocates note that the other people in every conversation never chose to be recorded. Mozilla Foundation’s Lauren Hendry Parsons told TechRadar that even with perfect data handling, the risk shifts to consumers rather than companies.
This is the same tension consumer BCI makers face when licensing sensing tech to wearable brands — I covered a similar pattern in Neurable licensing its non-invasive BCI to wearable makers. Whoever owns the sensor, the bystander problem stays unsolved.
Why builders should watch this
Bee sits in a growing category. Plaud has the NotePin S and Plaid One AI earbuds; startup Vibe ships Dot, a clip-on workplace device. As SiliconANGLE observes, AI wearables that listen — and sometimes see — are spreading from enterprise into personal life.
The open question for Amazon is whether ambient listening feels like a helpful memory or like Google Glass all over again. The practical takeaway: if your product captures passive context, give users visible state, cheap defaults, and a clear line separating personal utility from commercial action. Bee’s opt-in shopping boundary is a reasonable template — whether users trust it at scale is the part no spec sheet can settle.