The next version of the AI assistant may not wait for a wake word. New AI wearables are being built to sit on the body, listen throughout the day and convert ordinary conversations into memory, summaries and tasks.
Bee AI’s Pioneer wearable and Omi’s newly unveiled device show where this category is heading. The hardware can look simple, even cheap, but the real product is software that interprets the user’s day through large language models.
The shift from voice assistant to passive listener
Voice assistants have already moved from speakers and phones to wrists and faces. The important difference with this new class of AI wearable is that it can work in the background without the same kind of active prompt.
Bee AI’s yellow Pioneer bracelet recorded what was happening around the wearer during a day at CES. It did not store audio like a standard recorder app, according to the source article, but it processed conversations and produced personalized to-do lists and readable summaries of in-person chats.
Omi is pursuing a similar idea. Its wearable can record the environment to build an activity log, then use AI to turn that information into insights and tasks. It can be worn around the neck, but the company says it is best worn stuck to the forehead near the temple.
Omi’s device includes an electroencephalogram, and Omi claims that if a user thinks specifically about talking to the wearable, the device can understand that intention and prepare to receive the request. That is a more unusual interaction model than pressing a button or saying a wake phrase.
Bee AI shows the practical appeal
Bee AI was founded by Maria de Lourdes Zollo and Ethan Sutin. Both previously worked at Squad, where Sutin was the founder. Squad enabled media screen sharing in video chats so people could watch the same movie or YouTube video together remotely.
Squad was acquired by X, when it was still called Twitter, and both founders briefly worked on Twitter Spaces. Zollo has also worked at Tencent and Musical.ly, which later became TikTok.
Sutin says he explored a personal AI assistant idea in 2016, when chatbots were attracting attention, but the technology was not ready. Bee AI launched its platform last February in beta and built an active feedback community before the Pioneer hardware began selling a little more than a week ago.
The Bee AI hardware is not required to use the service. A user can interact with Bee AI through the iPhone app, while an Android app is expected at the end of the month. Zollo says the wearable gives a richer experience because it can record continuously all day.
The device has two microphones for noise isolation and can be worn either as a wrist band or clipped to a shirt. Sutin says that if the wearer can hear another person in a busy environment, the wearable should be able to hear both sides of the conversation as well.
Tasks, memory and summaries are the selling points
The most obvious use case is for people who spend much of the day speaking with others. Sutin describes Bee AI’s target demographic as people “who talk a lot for a living.”
The app can show a summary of conversations from the day. At the end of the day, it creates a short account of what the day was like and can show the locations where conversations happened on a map.
Its central feature is the “To-Dos” tab. These tasks are generated automatically from conversations. In one example from the source article, a conversation with an editor about taking a product photo led Bee AI to create the task “Remember to take a picture for Mike.”
That capability can be useful, but it is not consistently accurate. The source notes that many generated to-dos may not actually be things the user needs to do. When the system does capture a useful task correctly, the experience can feel powerful.
Bee AI can also be connected to Gmail, Google Calendar and Google Contacts. Those integrations are meant to let users ask about an email or what comes next on a calendar, though the source article says that functionality was not available to try.
- Bee AI’s Pioneer wearable costs $50.
- Omi’s stick-on bead costs $89.
- Bee AI offers basic memory recall and summarization with the hardware.
- Additional features, including third-party app integrations, cost $12 per month.
The privacy issue is built into the product
Always-listening AI wearables are useful because they capture context. That is also what makes them sensitive. The device is not only interpreting the wearer’s words; it may process what nearby people say as well.
Bee AI’s Pioneer has an “Action” button. Pressing it once mutes the microphones, and pressing it again turns them back on. Pressing and holding the button can trigger user-configurable actions, such as processing the current conversation or waking the “Buzz” AI assistant.
There is no speaker on the Bee AI wearable, so spoken answers come through the phone. When the microphone is muted, a red LED appears. But when it is recording, there is no visible green LED showing that the device is picking up audio from the surrounding environment.
Zollo says a constantly active green LED would affect the wearable’s purported seven-day battery life. Still, the absence of a clear recording indicator creates a difficult consent problem, especially because recording laws vary from state to state in the US.
The company says the wearable is not storing audio, but users can view full transcripts of conversations, even when those transcripts are not always completely accurate. Sutin says everything captured is “treated as maximally sensitive,” that the business model does not include monetizing collected data, and that nothing will be shared with third parties. He also says no human can see the data.
Cloud processing remains the tradeoff
Bee AI does not process conversations locally on the phone. Sutin says edge processing is getting closer, but battery life remains a fundamental issue. For now, the data is processed in the cloud.
Bee AI uses different large language models depending on the task. The source article says the mix includes commercial and open source models, including OpenAI’s ChatGPT and Google’s Gemini, along with models hosted by Bee AI itself.
That architecture explains why the hardware can be relatively inexpensive while the service model matters. The bracelet or bead is only the sensor. The assistant lives in the processing pipeline that turns continuous speech into structured information.
The larger question is whether people will accept assistants that are most useful when they are most present. Bee AI and Omi point toward AI that does less waiting and more observing. For workers who live in meetings, calls and conversations, that could make daily memory easier to search. For everyone else in the room, it may be harder to know when a casual conversation has become machine-readable data.