AI驱动的语音革命:Otter.ai的创业之路与企业智能未来 EO 2026-03-24

教育的变革与历史语音知识的遗失

两天前,我还在哈佛大学。我发现许多教授实际上不允许学生使用像 Otter 这样的工具来辅助学习。我认为这是一种过时的思维模式。我们当前的教育体系,其根基至少可以追溯到 100 年前。在信息爆炸和技术飞速发展的今天,教育机构应该允许学生利用一切可用的 AI 工具来提升学习效率。尽管人类彼此交流已有数百年、数千年之久,但历史上绝大多数语音知识都已随着时间的流逝而消散。我们从未听过莎士比亚或查尔斯·达尔文本人的声音,这无疑是人类知识和智能的巨大损失。正是基于这种思考和对现状的洞察,我们预见到语音 AI 在未来将扮演极其重要的角色。我是 Sam,About AI 的联合创始人兼 CEO。我们最初以一款转录工具起步,随后将其发展为 AI 会议助手,现在我们正在构建一个以会议为中心的企业知识库,并在其上集成 Agentica 工作流。迄今为止,我们已服务超过 3500 万用户,年经常性收入(ARR)突破 1 亿美元。如今,越来越多的企业正积极采用我们的产品来管理和利用海量的会议内容。

Original English

I was at Harvard University just two days ago and a lot of professors actually don't allow students to use a tool like Otter to help them learn. I think that's that's old thinking. The way we do education, the system was created at least 100 years ago. They have to allow the students to use whatever AI tools. Although human beings have been talking with each other for hundreds, thousands of years, most of the voice knowledge in the history has been lost. We never heard from Shakespeare. We never heard from Charles Darwin. That's tremendous loss of human knowledge and human intelligence. With those frustration and insight, we thought that in the voice AI will be really huge in the future. I'm Sam. I'm co-founder and CEO of a about AI. We started as a a transcription tool, then evolve it into a AI meeting assistant and now we're building a meeting centric enterprise knowledge base with Agentica workflows on top of it. So far we have over 35 million users. We exceeded $100 million in ARR. Now enterprises are adopting it to manage their huge meeting content.

创业启程:从斯坦福到 Google,再到改变会议记录

我在斯坦福大学完成了博士学位,并从我的博士导师 大卫·谢尔顿(David Sherton)那里学到了很多。他具备识别能够产生巨大影响、改变世界事物的远见。这也是为什么在 拉里·佩奇(Larry Page)和 谢尔盖·布林(Sergey Brin)尚无成就之时,他便看到了他们的才华,并写下了 10 万美元的支票。我从他身上学到了许多关于“宏大思考”的经验。我在 2006 年至 2010 年间曾供职于 Google,担任 Google 地图定位平台的负责人。之后,我在 2010 年离开了 Google,并在 Palo Alto 创办了一家移动初创公司。我们是第一家构建位置追踪系统,并能在移动设备上进行持久化感知的公司,能够深入理解用户的移动行为,从而为他们提供更个性化的移动服务。这家公司后来成功被收购。

Original English

I did my PhD at Stanford University and I learned a lot from my PhD advisor. His name is David Sherton. He has the vision about what will generate a big impact, what will change the world. That's why he actually recognized the talent of Larry Page and Sergey when he wrote a $100,000 check to them before they had anything. I learned a lot from him in terms of thinking big. I was at Google 2006 to 2010. I was the lead of Google map location platform. Then I quit Google in 2010 to start a mobile startup in Palo Alto. We were the first that build a location tracking system and also do persistent sensing on mobile devices that understand users mobile behaviors so that we can personalize the more mobile services for them. That company was successfully acquired.

解决痛点:会议记忆的挑战与 Otter.ai 的诞生

随后,在 2016 年,我开始构思一些新的、更具颠覆性的事业。在我创业初期,我需要与投资者、内部团队和客户进行大量的会议。然而,我发现自己很难记住所有会议的关键内容,同样难以将这些宝贵的知识有效地分享给团队成员。因此,我坚信一定存在一种更好的方式来解决这个普遍存在的痛点。

回到 2016 年,我们提出了一个大胆的想法:记录下所有的会议内容,并使其能够轻松地与团队成员共享。然而,这两个核心功能在当时都让大多数人感到不适。首先,被录音本身就令人感到不自在;其次,将会议笔记公开分享给他人也是一种非常规的做法,因为在传统观念中,人们习惯于使用纸质笔记本进行个人化的笔记记录。我们预见到,随着技术的发展,这种观念和文化必将发生转变。因此,我们致力于构建一款能够驱动并促进这种变革的产品。你可以说服一部分人,但无法强求所有人接受。对于任何新产品而言,这都是一个必然经历的采纳过程。那些早期采用类似产品(例如自动驾驶技术)的用户,能够更快地体验到其价值和优势,从而变得更加高效、更有生产力,并能向同事展示其带来的实际效益,进而帮助说服更多用户。因此,你需要选择一个大多数人尚未被说服,但潜力巨大的方向。

Original English

Then in 2016 I was thinking about something new and something bigger. While I was building the first startup, I had a lot of meetings with investors, a lot of meetings with our internal team and customers. Really hard for me to remember all the meeting content. It's also hard to share that knowledge with all the team members. So, I think there must be a better way to address that. Back in 2016, we say we're going to record everything. We're going to enable it to be shared with other team members. Both made most people uncomfortable. Number one, being recorded is uncomfortable and also share meeting notes with other people. It's uncommon because traditionally people take notes on the paper notebook. It's a personal thing. We anticipate that the mindset will change, the culture will change. So we build the product that enable that change. You can convince some people. You cannot convince everyone. That's okay. You know for any new product it it follows certain adoption curve. For people who adopted a product like auto early, they actually get value and benefit sooner. They can become uh more effective, more productive and they can show the value to their colleague. You know they can help convince the other users as well. So you have to pick something that most people haven't haven't been convinced yet.

深层技术是基石:构建差异化与未来智能接口

如果你希望一项事业能够真正取得巨大成功,就必须拥有深厚的技术根基。我本人就拥有技术背景,对技术和工程充满热爱。我深知,许多颠覆性的伟大公司,如 Google,都是建立在扎实的核心技术之上。如今,市面上有大量的 API 可以让你快速构建一个会议笔记记录工具。理论上,任何大学生在 2016 年(十年前)就能做到这一点。然而,如果我们当时仅仅依赖现有的 API,那么在我们决定自主研发语音识别技术时,我们可能会落后市场数年。我们并不知道这项技术需要多长时间才能成熟,也清楚其中蕴含的巨大风险。与 Google 或 Microsoft 等巨头相比,我们的资源、资金和人力都相对匮乏。如果它们在技术上迅速追赶上来,我们该如何应对?我们的战略选择是自主研发深层核心技术,以此为基础,我们在未来创造一场新的技术革命。对于一个新兴的初创公司而言,如何构建独特的差异化优势,始终是最大的挑战。从创立之初,我们就坚信自主掌握核心技术是必由之路,因为拥有自身技术意味着我们可以有效控制成本。相比之下,若依赖第三方 API,你不得不支付高昂的费用,这将极大地限制你提供免费服务的可能性。

目前,仍存在一些复杂的难题,亟需 AI 科学家们投入精力去攻克。例如,面对数亿条语音数据,我们该如何有效地利用它们来真实地模拟人类对话?我们又该如何精确地建模会议中多个发言者之间复杂的互动过程?这仍然是一个悬而未决的科学问题。要解决这些问题,仅仅依赖第三方 API 是远远不够的,你必须自主构建深层的 AI 技术栈。请记住,如果某个技术门槛对你而言过于轻易,那么对其他成百上千的竞争者来说,也同样会如此。

技术的革新必然会驱动行为的改变。回顾过去 50 年,在互联网尚未普及的年代,电子邮件似乎是唯一的通信方式。然而,随着 Slack 这样的新工具的出现并流行起来,人们开始减少使用电子邮件,转而依赖 Slack 进行沟通。同样,当语音技术日渐成熟,我们预测语音将成为企业智能(Enterprise Intelligence)的首要交互界面。在不久的将来,你可能不再需要进行大量的文字输入。人们将极少进行文字写作,也很少使用键盘输入信息。他们可以直接通过语音交流,因为说话比写作更直观、更便捷。他们只需开口,AI 便能自动将内容转化为文字。这一趋势已然显现,许多人已经开始利用 AI 来撰写文档、电子邮件和社交媒体帖子。这种改变正在加速发生。因此,我们的核心观点是:语音正逐步成为商业智能(Business Intelligence)的主要交互方式。展望未来十年,我们仍有大量的工作要做,市场渗透率可能仅达到 95% 甚至更高。我可以说,全球 90% 到 99% 的用户尚未采纳 Otter 这样的工具。我们必须以长远的眼光,着眼于未来十年,而非仅仅关注当下。这才是能够成就“时代性公司”的关键所在。

很多人将创业比作跑马拉松,但实际上,创业的艰难程度远超跑马拉松。我个人已成功完成了 11 次马拉松,并且即将在两个月后参加一次新的比赛。长跑经历无疑帮助我保持了健康,有效应对了巨大的压力,并克服了无数挑战。大多数人在面对困难时都会很快放弃。如果你正在构建一项极具挑战性的事业,那么遇到的困难往往是符合预期的。关键在于你必须持之以恒,不断地追求你的目标。

Original English

If you want to really go big, you need to have deep technology roots. I came from technology background. I like technologies. I like engineering. I also see that a lot of revolutionary companies are built on deep technologies like Google. Today there are a lot of APIs you can use to quickly build a meeting note taker. Any college students can do that already in 2016 10 years ago. At that time if we were waiting for someone else to create the API you know we we would be many years late when we decided to build our own speech recognition technology. We didn't know how long it would take. We know there there is a lot of risks. We know that we have a lot less resource, a lot less money, a lot less people than Google or Microsoft. What if Google or Microsoft or other people catch up fast? Our choice is to build deep technologies which can enable us to create a new revolution in the future. What differentiation can you create? That's the biggest problem for a new startup. From day one, we always had that belief that that should be the way that should be the right way because we're we own our own technology. So we can keep the cost low. If you use a third party API, you have to pay them a lot of money that limit how much free service you can provide. There are still deep problems that require uh AI scientists to work on. For example, you know, when we have hundreds of millions of voice data, how do we use that to truly model human conversation, how do we model the interactions of multiple speakers talking to each other in the meeting? That's still a unsolved problem. To solve that problem, you cannot just rely on third party APIs. You have to build your own deep AI tag. If it's too easy for you to build, it's very easy for 100 other people to build as well. Behavior always change when you have new technologies. If you look back in the last 50 years, right before internet became so common, it feels like we have been having emails forever. But then a new tool like Slack became popular. Then people actually send fewer email. They rely on Slack. But then you know with when the voice technology become much more mature. We think that voice will become the primary interface for enterprise intelligence. You probably don't need to write so much in a few years. People will rarely write anything. They will rarely use keyboard to write anything. They can just talk because talk is easier than writing. They can just talk and our AI will write everything for you. It's it's start to happen. A lot of people actually use AI to write documents, to write emails, to write linking post. It's already happening. It will only accelerate. So our view is that voice is becoming the primary interface of business intelligence. Looking forward to the next many years, there's still a long way to go. At least 95% or even higher. I would say 90 99% of the world hasn't adopted a tool like Otter yet. We have to look at the next 10 years not just today. That's how you know this generational companies are built. People say building a startup like running a marathon. Actually building a startup is way harder than running a marathon. I've run 11 marathons. I'm going to run another one in 2 months. That definitely helped me stay healthy, handle stress, help me push through all the challenges. Most people give up pretty fast. If you're building something challenging, the difficulties are as expected. You have to persist and and continue pursuing your goal.

📌 文中提及的人物和组织

人物: Sam Liang, David Sherton

公司/组织: Otter.ai, Google, Stanford University

产品/模型: Otter.ai, Agentica

关键字: voice-ai meeting-intelligence startup-journey enterprise-knowledge future-of-communication