Otter.ai

Sam Liang was the Google engineer who put the blue dot on Google Maps. He left Google, built a location-context startup called Alohar Mobile, sold it to Alibaba, and then sat down to solve the next hard problem: not where you are, but what you're saying. The insight driving AISense — the company he co-founded with Yun Fu in 2016 that became Otter.ai — was deceptively simple: every important decision in a business happens first in a spoken conversation, and almost none of that intelligence is ever captured, searchable, or usable afterward. Meetings are where strategy is made, where clients commit, where plans change. But the moment a meeting ends, that information evaporates — unless somebody took good enough notes, remembered the exact wording, or happened to be paying attention at the right moment. Otter was built to make that forgetting impossible.

The product they built is a real-time transcription engine backed by a proprietary speech model — not a third-party API — custom-tuned for meeting acoustics, accent variation, background noise, and overlapping speech. OtterPilot, the AI meeting bot, joins Zoom, Google Meet, and Microsoft Teams calls automatically, transcribes every word with speaker attribution, generates structured summaries with action items, and pushes follow-ups to Slack, HubSpot, Salesforce, or email without human intervention. Otter AI Chat lets users ask questions across every meeting in their history — "what did the CTO say about the product timeline last quarter?" answered in seconds from a corpus of transcripts that would take hours to search manually. In 2025, Otter crossed $100M ARR with fewer than 200 employees — over $500K in revenue per employee, an efficiency benchmark that rivals the best enterprise SaaS companies in the world. Thirty-five million users. One billion dollars in measurable customer ROI delivered. And Sam Liang still answers his own LinkedIn comments.

The pivot that defines Otter's next chapter is not transcription — it is agency. In October 2025, Otter launched its enterprise knowledge base product alongside a suite of domain-specific AI agents: a Sales Agent that joins calls, syncs CRM data, and drafts follow-up emails autonomously; a Recruiting Agent that transcribes interviews and extracts structured candidate assessments directly into ATS; and an Education Agent that supports students and teachers with lecture notes, quiz generation, and concept summaries. An Otter API and MCP server now allow enterprise teams to connect Otter's decade of accumulated meeting intelligence to external AI models — Claude, ChatGPT, and others — turning the meeting corpus into a live knowledge layer that any agent can query. Liang's thesis for 2026: voice is AI's most underutilized input modality, and the organizations that treat their conversation history as a strategic asset — searchable, structured, agent-accessible — will out-execute those that don't. "In 2026, the dominant system of record won't be your wiki or your CRM," he wrote publicly. "It will be your conversations."