The best knowledge in your company isn't in the wiki—it's trapped in the 47 meetings you attended this month that nobody else can search.
The Summary
- Otter is launching AI features that turn meeting transcripts into searchable institutional knowledge, integrating with Google Drive, Jira, Salesforce, and Notion to pull live data when answering questions
- The system now supports the Model Context Protocol (MCP), letting ChatGPT and Claude access Otter's meeting data with permission
- CEO Sam Liang's key insight: written documents go stale fast, but the latest company intel lives in meetings that aren't systematically searchable
The Signal
Most corporate knowledge systems have an expiration date problem. You write the doc, it's perfect for two weeks, then reality shifts and nobody updates it. Meanwhile, the actual decisions, the real context, the stuff that matters gets discussed in meetings that vanish into the void the moment they end.
Otter's CEO Sam Liang frames this as the fundamental knowledge management problem: documents become obsolete fast, updates happen in meetings, and research shows white-collar workers spend huge chunks of time in those meetings. But meeting knowledge traditionally stays locked in individual accounts or never gets captured systematically at all.
"People create documents, but documents become obsolete really fast."
The new Otter features attack this from three angles:
- Cross-platform integration with Google Drive, Jira, Salesforce, and Notion for pulling live data
- MCP support so other AI agents can query Otter's meeting database
- Enhanced chat that lets users specify which meetings to search and combine meeting insights with other data sources
The MCP piece is where this gets interesting for the agent economy. Model Context Protocol is becoming the standard way for AI systems to share context. By supporting it, Otter isn't just building a better meeting tool. It's positioning meeting transcripts as a queryable data layer that any compliant AI agent can access.
Think about what that enables. An AI agent working on a sales proposal could query relevant customer meetings, pull the latest pricing discussion from last week's product sync, and cross-reference both against live Salesforce data. All without a human manually digging through transcripts or Slack threads.
The company already has what it calls channels—essentially shared repositories for meeting transcripts from specific teams or projects. This new functionality turns those channels into something closer to living databases that update themselves every time someone talks.
The Implication
If you're building AI agents for enterprise work, meeting transcripts are no longer just nice-to-have context. They're becoming required infrastructure. The actual state of a company's decisions, priorities, and knowledge increasingly lives in conversations, not documents.
For companies, this creates a forcing function. Do you let AI systems access meeting data broadly, or keep it siloed? The permission question isn't technical anymore. It's cultural. And it's going to surface fast once agents start asking for access to conversations they weren't invited to.