Our Vision
Conversations are where decisions live.
We make them persistent.
The most consequential information in any organization is exchanged verbally. Pipeline reviews. Client calls. Product debates. Board discussions. Yet virtually none of it is captured with the structure, attribution, or connectivity required to act on it at scale. We are building the infrastructure layer that closes this gap.
The problem
Current meeting capture tools fail at six levels. Each failure compounds downstream.
Observer effect
Recording bots join as visible meeting participants. Prospects self-censor. Clients question data handling. Enterprise IT blocks the bot outright. The capture mechanism distorts the very conversation it is meant to preserve.
Broken attribution
A transcript labeled “Speaker 1” and “Speaker 2” is a log, not intelligence. It cannot drive CRM updates, inform deal reviews, or support account handoffs. Attribution is what makes meeting data actionable.
Platform fragmentation
Organizations run meetings across Google Meet, Zoom, and Teams, often in the same deal cycle. Each platform captures data in its own silo. There is no unified archive, no cross-platform search, no way to trace a relationship across the dozen meetings where it actually developed.
Single-meeting blindness
Native summaries treat each meeting as an isolated event. The most valuable signals, a prospect's shifting sentiment, recurring objections across a pipeline, a champion's move from advocacy to hesitation, only emerge across meetings. No single-meeting tool can surface them.
Knowledge walks out the door
When a senior AE leaves, the relationship context, deal nuances, and verbal commitments leave with them. The new rep inherits a CRM record and a few bullet points. Meeting intelligence should be organizational memory, not personal memory.
Dead-end outputs
Platform-native summaries are terminal. They cannot be queried, enriched, routed to CRMs, or consumed by AI agents. Meeting data becomes infrastructure only when it is structured, attributed, and accessible via API.
First principles
These are not features. They are architectural commitments that constrain every product decision.
Observation without interference
If the capture mechanism changes the conversation, every downstream output is built on a distorted signal. IceCubes operates as a browser extension within your own session. No bot joins. No participant appears. The conversation remains unaltered.
Identity as a first-class construct
Attribution transforms text into intelligence. The difference between “someone raised a SOC 2 concern” and “the VP of Engineering raised a SOC 2 concern” is the difference between noise and signal. IceCubes reads speaker names directly from the meeting platform UI. No voice training. No manual correction.
Structure over summaries
Summaries reduce information. Structure preserves it. A well-structured meeting output is a set of discrete, addressable objects: action items with owners, qualification signals with evidence, objections with attribution. Each object can be routed to a CRM, a task manager, or an AI agent.
Connected by default
Meeting intelligence that lives in a standalone app decays. Value is inversely proportional to the steps required to act on it. IceCubes syncs to HubSpot and Salesforce, posts to Slack, triggers Zapier workflows, and exposes a REST API with an MCP server for AI tools.
How it comes together
These principles manifest across three phases of every meeting.
During the meeting
- Real-time transcript with speaker names from the platform UI
- Lightweight sidebar (Meet) or floating widget (Zoom/Teams)
- One-click screenshots for visual context
- Google Meet, Zoom, and Microsoft Teams
After the meeting
- AI summaries from 30+ templates or custom formats
- Action items with assignees and due dates
- MEDDIC, BANT, objections, competitor mentions, pricing signals
- Smart Tags for custom meeting classification
- IceCubesGPT (up to 15 meetings)
Across the organization
- CRM sync to HubSpot and Salesforce
- Team sharing with granular permissions
- Searchable archive across all platforms
- REST API and MCP server for AI tools
- Slack, Zapier, and webhook integrations
Where this leads
We are not building a better note-taking app. We are building toward a state in which the information exchanged in meetings flows automatically into the systems where it is needed, with the structure and attribution required for each system to act on it without human intermediation.
A sales call populates qualification fields in Salesforce before the rep opens the browser. A product discussion creates tickets with full context before the PM opens Jira. A coaching review surfaces patterns across fifty conversations that no individual manager could track manually.
This is not about replacing human judgment. It is about ensuring that judgment is informed by complete, accurate, and well-structured information from the conversations where it was first expressed.
See the output for yourself
Install the extension. Join a meeting. Review the transcript, summary, and structured insights. AI summaries are free on all plans, no credit card required. Insights and action items use AI credits.
Read our blog
Add to Chrome
Add to Edge