Google AI Edge Foresight: What Is the Offline AI Meeting Assistant?

Google Introduces AI Edge Foresight, an Offline AI Meeting Assistant

Google has introduced a new experimental application for users who want to manage meeting notes with artificial intelligence. Called Google AI Edge Foresight, the Mac application can transcribe meetings, enrich notes and help users search across past conversations and documents with AI assistance.
One of Foresight’s most notable features is its ability to perform all of these tasks directly on the device without requiring an internet connection.

Meeting Notes Are Processed Without Being Sent to the Cloud

Google AI Edge Foresight works directly with a computer’s system audio and microphone. This allows it to follow conversations across different meeting applications without being tied to a specific video conferencing platform.
In addition to transcribing conversations during meetings, the application can supplement a user’s brief notes with information discussed during the session. This means users can focus on capturing key points instead of manually documenting the entire meeting.
According to Google, processing data directly on the device offers an important advantage in terms of privacy. Meeting recordings and personal documents do not need to be sent to the cloud for processing. The application’s core features can also continue to function when there is no internet connection.

Users Can Ask Questions About Meetings with Gemma 4

Foresight is not limited to converting speech into text. The application includes an AI assistant powered by Gemma 4 that can answer users’ questions based on the content discussed during a meeting.
For example, after a lengthy meeting, users can ask what decision was made, which task was assigned to whom or how a particular topic was discussed using natural language.
The application also allows users to view the full meeting transcript.

Documents Can Also Be Added to a Personal Knowledge Base

One of AI Edge Foresight’s key features is that it can work not only with meeting recordings but also with users’ own files.
PDFs, documents, meeting transcripts, notes and different types of media can be turned into searchable sources of information within the application. Users can then ask questions in natural language to find the information they need across these materials.
Google uses the EmbeddingGemma 2 model in this system to map different types of data, including text, audio, images and video, into a shared representation space. The model has 740 million parameters and is designed to run directly on the device.
This approach allows users to find information from a previously uploaded document during a meeting or quickly revisit a topic discussed in an earlier meeting.

Optimized for Apple Silicon

AI Edge Foresight is currently positioned as an experimental application for Mac computers and is optimized to run on Apple Silicon processors.
By using local AI models, Google aims to provide a meeting assistant experience that does not depend on a constant internet connection. This approach can also reduce the connectivity and data transfer requirements typically associated with cloud-based AI services.

A New Competitive Area for Meeting Assistants: On-Device AI

AI-powered meeting tools have become increasingly common in recent years. Google already offers features through Gemini that can summarize Google Meet conversations and generate action items. Foresight takes a different approach by focusing on performing AI processing largely on the user’s own computer.
This model may be particularly useful for confidential business meetings, sensitive project documents or work environments with limited internet access.
Although AI Edge Foresight is still an experimental project, Google’s investment in on-device artificial intelligence suggests that privacy, offline access and personal knowledge search could become increasingly important features in the future of AI-powered meeting assistants.