The AI that joins your Zoom call, transcribes every word, summarizes the chaos, and - lately - answers questions out loud while the meeting is still happening.
It is Tuesday morning, 10:03 a.m. Pacific. Somewhere in California, six product managers click into a Zoom call. One of them is late. Another forgot to mute the dog. None of them open a notebook. None of them open a Google Doc. They don't have to. Sitting silently in the call - a polite little participant tile - is an assistant called Otter. It is already listening, already transcribing, already deciding what counts as an action item.
This is the unglamorous miracle Otter.ai has been quietly engineering for nearly a decade. Not artificial general intelligence. Not a chatbot in a search box. Just the slow, deliberate annihilation of the meeting recap email.
For most of the last decade, conversations at work have been treated as a kind of disposable currency. They happen, decisions get made, then everyone scatters back to Slack to relitigate what they thought they heard. The follow-up email - that small, dreaded artifact - was the only bridge between what was said and what would actually get done.
Sam Liang, the engineer who would later co-found Otter.ai, found this absurd. He had spent his Google years building the blue dot that anchored you to your spot on Google Maps. Now he watched billion-dollar decisions evaporate the moment a meeting ended. It is a particularly modern irony: we record everything we type, log every click, A/B test every button - and the most important room in the building, the one where humans are actually deciding things, leaves no transcript at all.
In a 2016 industry survey, 73% of professionals admitted to taking no formal notes during recurring meetings. Most relied on memory. Memory, as it happens, is not very reliable.
Sam Liang and Yun Fu started the company in 2016 under the name AISense. Liang had already had one quiet hit - founding Alohar Mobile, a context-aware location platform that Alibaba scooped up in 2013. Fu, his co-founder, brought deep distributed systems experience and a stubborn belief that real-time speech recognition could finally be made to work at scale, on commodity hardware, in noisy rooms full of people interrupting each other.
The bet was simple, and at the time, slightly unfashionable. While the rest of the AI world chased autonomous cars and image generators, Liang and Fu pointed at the unsexy middle of the workday - the calls, the syncs, the standups - and said: that. That is the dataset nobody has indexed yet.
Draper Associates wrote a seed check in late 2016. Horizons Ventures led the Series A a year later. In 2021, Spectrum Equity put $50 million behind a Series B at a moment when much of the AI press cycle was still obsessing over chatbots that hallucinated their way through poetry. Otter, meanwhile, was learning to spell people's names.
At a basic level, the product is straightforward: connect Otter to your Google or Microsoft calendar, and the assistant will show up to your Zoom, Teams, or Google Meet calls without being invited twice. It records. It transcribes. It identifies speakers. It generates a structured summary with action items, key points, and - when slides get shared - the slides themselves, dropped neatly inline.
Joins automatically. Now voice-activated - ask it a question during the call and it answers in the chat.
A conversational layer that lets you query a single meeting or every meeting you've ever had.
Drafts follow-ups, flags buying signals, syncs to Salesforce and HubSpot before the next coffee.
Live captions and lecture notes built for classrooms. Custom vocabulary handles the jargon professors love.
The version that shipped in 2024 - the Otter Meeting Agent - was the moment the product crossed a line. Previously, Otter was a stenographer. Now it is something closer to a colleague who occasionally chimes in. You can ask it, mid-meeting, what the last quarter's revenue was, or what was decided about the launch date, and it will return an answer drawn from your prior transcripts. It is the kind of feature that sounds gimmicky in a press release and turns out to be addictive in practice.
Otter does not tend to brag, which makes the numbers more interesting when you bother to look them up. Public estimates put annual revenue around the $100M mark. Headcount sits at roughly 200. Customers span the obvious sales and education verticals and the less-obvious ones too - journalism, podcasting, legal intake, market research, a surprising amount of accounting.
The partner roster reads like a who's who of the modern workday. Zoom. Microsoft Teams. Google Meet. Salesforce. HubSpot. Dropbox. Each integration is small on its own and quietly load-bearing in aggregate. Otter is the kind of company that wins by being everywhere you already are, which is much harder than it sounds.
Ask Liang what Otter is for, and the answer drifts toward something philosophical: he wants the spoken word to carry the same weight at work as the written word. Documents get versioned, searched, linked, and audited. Conversations get forgotten. Otter is the wedge that levels those two assets, by turning every meeting into a permanent, queryable record that any colleague - or any algorithm - can pull from later.
It is, in its way, a deeply democratic project. The people who currently win at meetings are the people with the best memory, the loudest voice, and the most patient note-taker on their team. Otter quietly reshuffles those advantages. If everyone gets the same transcript, the same summary, the same searchable archive, then meetings stop rewarding the people who happen to take the best notes and start rewarding the people who say the best things.
For three years, the entire AI conversation has been about generation - what models can produce. Otter is part of a quieter and arguably more durable thesis: that the bigger prize lies in capture. The conversations happening in offices, on Zooms, in classrooms, at podiums - that is the unstructured data nobody has indexed yet, and it is roughly where the world's knowledge actually lives.
If Otter is right, the next decade of enterprise software will not be defined by the chatbot you talk to. It will be defined by the assistant that has been listening the whole time, and now finally has something useful to say about it.
Back to that Tuesday morning. The meeting wraps at 10:47. The product managers say goodbye, click leave, scatter. Two minutes later, every one of them has a summary in their inbox. The action items are bulleted. The decisions are tagged with the speaker who made them. The slide from minute twelve is right where it should be. Nobody had to write a recap. Nobody had to remember anything. The conversation, for once, survived the conversation.
That is the Otter trick. It looks like a small one. It is not.