Google releases a new local-first Granola competitor

Editor's note: Meeting notes are becoming one of the most crowded arenas for generative AI deployment. From Otter.ai to Fireflies.ai, from Microsoft Copilot to Zoom AI Companion, nearly every collaboration platform is making automatic note-taking a standard feature. Now Google is entering with a new app called AI Edge Foresight, choosing a more differentiated route: local-first and offline-capable.

According to TechCrunch, Google's newly released AI Edge Foresight app directly targets the recently closely watched note-taking tool Granola. It can transcribe meeting conversations, generate notes, and answer questions without relying on the cloud, with all core computation performed on-device. This means users can get an AI meeting assistant experience without uploading sensitive audio to servers.

Local-First: A New Selling Point for Privacy and Efficiency

"Local-first" is not a new concept, but it has special significance in meeting note-taking. Meeting content often contains trade secrets, product roadmaps, personnel discussions, and even legal information, and enterprise customers are extremely sensitive to data leakage. Although traditional cloud solutions have stronger model capabilities, they require audio to be transmitted, stored, and processed on third-party servers, which often makes compliance teams hesitate. AI Edge Foresight puts transcription and inference on-device, significantly reducing such concerns.

Beyond privacy, local processing can also deliver lower latency and zero marginal API cost. After users speak, transcription and summaries are generated almost instantly, without waiting for a network round trip. For users who travel frequently, have unstable networks, or are in airplane mode, offline capability is also a practical draw. Google's move follows the broader trend of on-device AI: Apple Intelligence, Copilot+ PCs, and NPUs from Qualcomm and MediaTek are all driving large models down from the cloud to phones and PCs.

TechCrunch notes that Google's new app AI Edge Foresight directly targets Granola, offering offline meeting notes, transcribing conversations, generating notes, and using on-device AI to answer questions.

Why Granola Became the Target

Granola is one of the fastest-rising AI note-taking tools of the past two years. It entered through Mac, emphasizing lightweight, fast operation that does not interrupt the meeting flow, allowing users to add their own notes during meetings, after which AI merges the transcript with human notes to generate structured minutes. This "human-AI collaboration" experience made it quickly popular among entrepreneurs and investors, and also proved that meeting notes are not just simple transcription but a comprehensive need for context, action items, and knowledge management.

Google's choice to target Granola shows that it is after not basic transcription but higher-level meeting intelligence. If AI Edge Foresight can combine on-device models with the Google ecosystem, it may further connect Calendar, Contacts, Gmail, Google Meet, and Workspace documents. For example, it could automatically generate minutes after a meeting and link them to the calendar event, or infer project background from historical emails. This kind of ecosystem synergy is an advantage that standalone tools find hard to replicate.

The Real Challenges of On-Device AI

However, local-first does not come without costs. On-device models usually have fewer parameters and smaller context windows than large cloud models, so transcription and summary quality may suffer with overlapping speakers, technical jargon, accents, and long meetings. Battery life, heat, and memory usage are also issues that mobile devices must solve. If Google wants AI Edge Foresight to become an everyday tool, it will need to strike a delicate balance in model compression, quantization, and task scheduling.

Another challenge is cross-platform coverage. Granola was initially Mac-focused and is expanding to more platforms; Google's app may prioritize Android and ChromeOS before considering Windows and macOS. If it can only be used on some devices, it will limit network effects. In addition, the enterprise market also requires admin controls, compliance certifications, and integration with existing meeting systems, none of which a single-point app can quickly accomplish.

The Meeting Notes Arena Will Become More Crowded

Google's entry sends a clear signal: meeting notes are moving from a standalone SaaS feature to a foundational capability of platform-level AI assistants. Microsoft has embedded Copilot into Teams, Zoom has launched AI Companion, and Cisco and Slack are following suit. Standalone tools such as Granola, Otter.ai, and Fireflies.ai must defend their positions with more refined experiences, cross-platform support, and depth in vertical scenarios.

For users, intensifying competition means better privacy options and a lower barrier to use. If local-first AI meeting notes mature, they could push the entire industry to rethink data architecture: which inferences must go to the cloud, and which can be completed on-device. The more likely future form is a hybrid architecture—sensitive data processed locally, complex inference called to the cloud on demand—balancing privacy, cost, and capability.

Currently, Google has not disclosed the full features, supported platforms, and business model of AI Edge Foresight. But one thing is certain: as on-device model capabilities improve, the next round of competition in the meeting notes market will not just be about "who transcribes more accurately," but about "who can better understand the user's meeting context while protecting privacy."

This article was compiled from TechCrunch.