Otter.ai pitches 'Conversational Knowledge Engine' as the missing system of record

Otter.ai is repositioning from meeting assistant to enterprise knowledge layer. In an interview with UC Today, co-founder and CEO Sam Liang argued the AI meeting-assistant category has been stuck on the same first step for eight years (transcription, summary, light chat) and that the next leg is connecting those conversations into a structured graph that sits alongside CRM, HRIS and ERP as a system of record.

The product framing for this is what Otter is calling the Conversational Knowledge Engine.

What the product actually claims to do

Liang told UC Today that Otter has processed billions of meetings since creating the AI meeting assistant category eight years ago, but that the industry as a whole has not moved past step one. "Transcription, summary, a little bit of chat, but they're not really connecting the knowledge," he said.

The Conversational Knowledge Engine is positioned as that connective layer. Per the UC Today writeup, it aggregates meeting data across an organisation and builds what Liang calls a longitudinal knowledge graph, mapping entities (clients, projects, topics, people) and tracking who said what, who the subject-matter experts are, and how knowledge evolves over time. The pitch is less "better transcript" and more "queryable institutional memory."

Liang's framing of the market gap: enterprises have CRM for sales data, HRIS for HR data and ERP for financial data, but no system of record for conversational data. "Most people haven't realised how much business intelligence has been generated in meetings," he told UC Today, "and how much has been lost over the past 100 years."

The scale claim underpinning the bet: enterprise employees outside engineering now spend more than 50% of their time in meetings, according to Liang's interview. Otter does not disclose the methodology behind that figure in the piece, and readers should treat it as the CEO's framing rather than independently sourced data.

The governance layer

The permissioning model is the part most relevant to enterprise buyers. Per UC Today, the engine uses a permission structure modelled on Slack channels, letting organisations control who can access which slices of the conversational graph. That matters because the failure mode of any cross-meeting knowledge graph is the same failure mode as any shared drive: a junior analyst querying "what did the board say about layoffs" and getting an answer they were never cleared to see.

Whether Slack-style channel permissions are sufficient for that risk surface is a separate question. Slack's model works for messaging because the unit of access is the channel a user is explicitly added to. A knowledge graph that ingests every meeting an organisation runs has a much larger surface area, including meetings that participants may not have realised were being indexed for later querying by others.

On the lawsuits

The headline phrase, that recording-consent lawsuits are "part of doing business," is Liang's own. UC Today notes the comment in the context of Otter's exposure to wiretap and two-party-consent litigation, which has dogged the meeting-assistant category broadly. The framing is unusually direct for a CEO selling into regulated industries, and our read is that it signals confidence the cases are survivable rather than dismissal of the underlying compliance question. Buyers in regulated verticals (financial services, healthcare, legal) will want to see the consent flow before the marketing.

Otter has not, per the UC Today interview, disclosed pricing, customer count for the new engine, or a specific GA date for the broader Conversational Knowledge Engine rollout. Re-reporters should confirm those against Otter's own announcements before citing.