为何企业拒绝他们曾要求的AI：代理CEO埃利亚斯·托雷斯

过去，我会这样想，一旦我有什么东西加载到你的浏览器上，我就想收费。你知道吗？就像，一美元。我当时就想，太好了，我赚了一美元。

In the past, I would be like, the minute I had something that kinda, like, loaded on your browser, like, I wanted to charge you. You know? It's like, a dollar. I was like, yes. I got a dollar.

但这次，我不想要你的钱。我更想知道你是否对此着迷，它是否足够吸引人，以至于我愿意为你公司做任何事。很多创始人以为只要他们做出了产品就万事大吉了，对吧？人们太天真了。有些人年收入五千万美元，却还天真地以为那产品真的有效。

This time, I'm like, I don't want your money. It's like, I want to know that you're obsessed and that it's so sticky and that I'm going to do everything at your company. There's so many founders that they think that just because they built something that they got it, right? People are delusional. People have $50,000,000 in revenue and they're delusional that that product works.

你知道吗？这次我真的想做好。我想打造一个产品，让人们争先恐后地给我钱，并说，这真的独一无二。这就是我现在的旅程。

You know? This time around, I really wanna do it right. I wanna build something that people are, like, like, throwing money at me and saying, like, this is really there's nothing like it. This is this is my journey now.

今天我们非常高兴为大家带来新一期的训练数据节目。我们邀请到了埃利亚斯·托雷斯，一位我们相识超过十年、共同经历三家公司的老朋友。多年前，我们初次见面时，他正忙于重建HubSpot的核心平台。之后，我们又在Drift——他的第二家创业公司中合作。最近，我们又与埃利亚斯在他的第三家创业公司Agency展开了合作。

We're really excited to share today's episode of training data with you. Today, we have Elias Torres, someone we've known for more than a decade across three companies. We first met Elias when he was busy rebuilding the core platform for HubSpot many, many years ago. After that, we got into business with him with Drift, which is his second startup. And then more recently, we got into business with Elias on Agency, his third startup.

Agency正在重新定义客户体验。今天，我们将了解这意味着什么，以及如何在AI时代建立一家公司。希望大家喜欢。埃利亚斯，欢迎来到节目。

Agency is in the midst of redefining customer experience. And so today, we'll hear a bit about what that means and also how to build a company in a world of AI. We hope you enjoy. Elias, welcome to the show.

谢谢。我一生都在等待这一刻。我知道你

Thank you. I've been waiting this for this moment my whole life. I know you

确实如此。我知道你一直在等。我们原本准备了一份精彩的脚本，但过去几分钟里我们聊到了一些热门话题。不如就从这些热点开始？跟我们说说AI领域正在发生什么吧。

have. I know you have. You know, we had we had this wonderful script all planned, and then over the last few minutes, we got into some real hot takes. Why don't we just start with the hot takes? Tell us what's going on in the world of AI.

我认为每个人都在要求AI，都说自己想要，但当你真的带来时，他们却拒绝接受。就像弥赛亚降临了，大家却说‘不，我们不需要’。这是个非常有趣的时期，因为我们拥有人类有史以来最先进的技术。它如此智能、如此聪明、如此强大，但人们却要求它达到完美境界，而他们甚至无法用聊天提示提出一个基本问题。所以我们完全误解了这项技术及其能力。

I think that everybody's asking for AI, everybody says they want it, and when you come and bring it to them, they just reject it. It's like the Messiah has arrived and everybody's like, No, we don't want it. And I think it's a super interesting time, right, because we have the best technology mankind has ever created. It is so intelligent, it is so smart, it is so capable, But yet, people are demanding this level of perfection when they're not even able to use a chat prompt to ask a basic question. And so I think that we are completely misunderstanding the technology and its capabilities.

所以这是期望值不匹配的问题，并非

So it's an expectation mismatch, not necessarily

技术缺陷。某种程度上我们习惯了前AI时代的技术，它能从数据库中精确查找字符串，实现完美匹配

a technological flaw. Think in some ways we're used to pre AI technology, right, that it was able to look up a string from a database and it was a perfect lookup and a perfect match

而且

and

因此人们习惯了那种完美标准，当你带来任何可能的不同答案或解决方案时，我认为人们正在与这种不完美性作斗争。

so I think they're used to that level of perfection and so when you bring any possibility of a different answer or a different solution, I think people are struggling with that imperfection.

所以我们习惯了狭窄的功能范围和绝对精确。突然面对广泛功能但相对不精确时，人们很难适应这种转变。

So we're used to narrow functionality, perfect accuracy. All of a sudden we have broad functionality, imperfect accuracy, and people are having a hard time with that transition.

没错。我看到很多客户持半杯水心态。我对AI的态度是：如果它能完成比我原先多10%、20%、30%的工作，我就非常高兴了——这些本不需要我做，它节省了我的时间。但其他人认为，既然不是100%完美，我宁可什么都不要。

Yeah. I think I see a lot of customers being like half full, half empty. My approach to AI is like, if it's doing 10% more than I wasn't doing before, if it's doing 20%, 30%, I'm extremely happy. I don't have to do that, it saved me time. But other people, it's not 100% perfect, I'd rather have none.

我认为这种观点对人们来说是非常危险的。

And I think that's a very dangerous take for people to take.

实际上，我想到的例子是你之前和我们分享过的，关于你如何管理自己的邮件。对吧。你愿意分享那个例子吗？

And actually, the example that comes to mind is one that you shared with us before, which is around how you manage your own email. Correct. Do you care to share that example?

哪一个？我那个...让我看看我是否理解正确。是关于Agency的，对吧？Agency可以阅读我所有的邮件并创建草稿，我可以自动回复所有人。通常我在演示产品时，会直接向客户展示我的整个收件箱。

Which one? The one that I So let me see if I get this one right. So with agency, right? Agency can read all my email and creates drafts and automatically I can reply to everybody. So usually what I'm doing is when I'm demoing the product, I open my entire inbox to the customers right now.

就像，现在还不用担心，但我会想，哦，我刚和这个客户开了会，我们有这件事要跟进，我就直接点击发送按钮。是这个例子吗？

It's like, really not to worry about it yet, I guess, but I'm like, Oh, I just had this meeting with this customer and we had this thing and I'm going to send this follow-up. And I just hit the send button, you know? Is this the example?

是的，因为我们刚才在讨论邮件泛滥的问题已经难以应对了。很多人每天要花大量时间处理邮件。我理想中愿意付费的AI产品就是能帮我完全处理邮件的那个。

Yeah, because we were talking about how, like, the email deluge is just untenable at this point. Yeah. And like people spend hours and hours a day managing their email, or at least a lot of people do. Yeah. And like the AI product that I would love to pay for is one that could just take care of my email.

我们当时在讨论这个话题，你说你根本不用花时间处理邮件，实际上都是让Agency代劳的。

We were having this conversation. You're like, I don't spend any time managing my email. I actually just let agency do it for me.

完全正确。不过这不是重点

Absolutely. Which is not the point

代理权虽非唯一，但确是它能实现的功能之一。

of agency, but it's one of the things agency can do.

没错。我认为这是个很好的例子，它缩小了范围——我们为客户生成这些邮件和跟进内容。关键在于，如果你要管理500个潜在客户并试图与所有人沟通，哪种方式更好？是精心打磨两封完美邮件，还是给所有500位潜在客户都发送消息？对吧？

Yeah. I think that's a great example, I think that narrows it, right, is that we generate these emails for customers and follow ups. And the thing is that if you're managing, if you're trying to manage, you know, 500 pipeline and you're trying to communicate to everybody, what's best? That you deliver two perfect emails or that you send something to all 500 prospects? Right?

因为对方并不知道你花了多少时间。但就像轻推一下，直接发送那封跟进邮件。我的客户经常这样——我发出这些消息时几乎不看内容，有时可能觉得尴尬，但还是发了。部分原因可能是我现在的人生阶段，年纪大了就明白：永远说不出完美的话，发出就发出了。只要...你知道的，不会有大问题。

Because they individually don't know how much time you're spending on it. But just say ping and like send that email, right, to follow-up. And so what happens with my customers, right, is like I'm sending these messages and I kind of like almost don't I look at it, I might cringe sometimes, but I send it. Because partly it's this, I don't know, stage of my life, maybe because I'm old, I'm like, I already know that I'm never going to say anything perfect, so if it goes out, it goes out. And like, as long as, you know, it's going to be fine.

但问题在于，当我从用户体验角度观察客户使用产品的录像时，发现有人花了二十分钟却几乎没修改邮件。有时他们只改了称呼，或者觉得'这是我的日历链接'太强势，改成'告诉我您方便的时间'。二十分钟后邮件才发出。这就是当今企业难以扩展的根本原因——人为因素在中间制造了太多延迟、障碍和减速，尽管本意是为提升客户体验。

But the problem is I see customers, I watch them on recordings, right, of using the product from a user experience perspective and I'm like, Wow, this person just spent twenty minutes and they really didn't edit the email. Sometimes they're just changing the salutation or like they didn't like how it was too aggressive to say, Here's my calendar link. And they were like, Let me know what times. And then they're like, twenty minutes later an email gets sent. And so I think that that's fundamentally why businesses can't scale today, because there's humans in the middle, you know, to add so much delay and obstacles and slowdowns in the pursuit of delivering a customer experience, you know.

所以你是说人们害怕让AI掌控方向盘，但其实应该放手让AI来驾驶？

So you're saying people are scared to let AI take the wheel and they should just let AI take the wheel?

我是这么认为的。我们需要分清轻重缓急，就像单向门和双向门的区别。应该更宽容些——如果发出的邮件不够完美或有些臆想内容，我们双方都应该更包容地说'没关系，别担心'。发送方和接收方都需要更放松些。

I think so. I think we need to choose what really matters, know, like kind of one way doors, two way doors. We need to say like, is this going like we need to also have a little bit more leniency and forgiveness. Like if an email goes out, you know, that didn't say exactly the right thing or hallucinated, I think we all should be a little bit more forgiving and be like, it's okay, don't worry, And remove that fear from people. I think both sides, the sending and the receiving, needs to be a little bit more chill.

是的。我打算...

Yeah. I'm going to

下次我说些摘掉他最后一顶帽子的疯话时就用这个借口。抱歉，这是AI说的。

use that excuse next time I say some crazy shit that picks his last hat. Sorry. This is the AI.

但这不就是我们平常干的事吗？对啊。我们先说疯话，然后就像那个在美国公开赛上抢小孩帽子的家伙。

But isn't that what we do normally? Yeah. We say crazy shit, then it's like, like, the guy that stole the hat from the kid, you know, at at the US Open.

天啊没错，完全正确。

Oh, my god. Yes. Exactly.

先来段疯话，再来段更疯的。他完全可以甩锅给AI嘛，反正也没差，懂我意思吧？

First says some crazy shit, then he says some other crazy shit. It's a he could just throw an AI too. Might as well. You know?

呃，这让我想起

Well, it reminds

虽然我们可能会被他起诉

me Although, we might be sued by him

现在。你都没具体说是谁，所以我觉得没问题。通常在训练数据里我们不会涉及个人故事，但我觉得你的个人经历确实很特别很独特。能不能简单说说童年经历如何塑造了你现在的企业家特质？

right now. You didn't specify which guy, so I think we're good. Normally on on training data, we don't really get into personal stories, but I actually think your personal story is pretty exceptional and pretty unique. And so could you maybe just say a couple words about how your childhood shaped the sort of entrepreneur that you have become?

是的，完全正确。乔治，我觉得我自称工程师只是图个名头。我以前写过代码。但咱们聊聊个人经历吧，绝对的。我想我是在尼加拉瓜长大的，可以说是在我们所谓的共产主义时期，对吧？

Yeah, absolutely. I think that I am an engineer, George, just for the sake of it. I've written code before. But let's talk about the personal stuff, absolutely. I think I grew up in Nicaragua, in, I would say, what we call a communist era, right?

所以，那是个非常落后的第三世界国家，懂吧，我现在仍有文化冲击。我去别人家里——不是擅闯那种，是受邀去做客——还是会下意识确认。这让人很困惑。

So, very little, third world country, you know, I still have a shock. I go into people's homes, not that I go without, like, I do like they invite me, you know, just to make sure. Get confused.

你又不是在非法闯入。

You're not breaking in.

当然不是非法闯入。但当我打开美国人家的冰箱，里面塞满食物，每个抽屉都是成堆的盒装零食。要是在尼加拉瓜我家，所有柜子都空荡荡的，可能就剩几口锅碗瓢盆。

You're not breaking in here. But then I go in and I open the fridge and it's just like, just full of food. Americans, it's just like boxes and boxes and snacks and more, every drawer you open. If you came to my house in Nicaragua, everything was empty. Like, the only things that would be in there is like a few pots and pans.

我小时候真切记得，我们常对着空橱柜说'根本没吃的'。每顿饭都是现做现吃，就像及时生产模式。那种物资匮乏深入骨髓。有次我问妈妈要东西，她说'但凡能给，我怎么会不给？'

Like, I literally remember this as a kid, we were like, There's absolutely nothing to eat. There would be like, you know, just the food that you were going to make for that meal kind of thing. It's like, just in time. And so that's just a lot of scarcity, right? And so you grew up with I remember telling my mother, I would ask for things and one time my mom said something like, I would give you anything could if I'm not giving it to because I can't.

后来我学会问'妈妈，如果你买得起这个礼物的话...'。这和第一世界国家美国——所谓全球最伟大国家——形成鲜明对比。但看看当今我们的生活，至少我的圈子，简直走向了另一个极端。

And so I learned to ask, like, Mom, if you can buy me this gift, you know, if you can buy me this So it's just like a completely different world than the world that we live in. In a first world country, America, you know, greatest country in the world, you know? But then it's like, and then how we're living today, right? And a lot of people are like, at least the world that I'm living in today is just like the extreme opposite of that. Yeah.

你进IBM前的第一份工作是什么？

What was your first job? Pre IBM?

我是说，你真正意义上的第一份工作，任何能拿到报酬的工作。我觉得这里有个比较。

I mean, your actual first job that you got paid any sort of money for, for doing any sort of work. I think there's a competition.

听起来更体面的是我和母亲一起打扫办公室的经历。我们两样都做过。刚来这个国家时，我同时在麦当劳打工和做办公室清洁，那时我第一次领到薪水，当时17岁。

The one that sounds better is the one I was cleaning the offices with my mother. We did both. I was working at McDonald's and cleaning offices when we first came to this country I got paid. I was 17 years old.

我记得你提过你母亲有台笔记本电脑，你因此成了她的技术支持，这就是你接触科技的契机。

And I remember you told me your mom had a laptop and you became tech support and this is how you got introduced to technology.

那是在来美国之前。我和父亲生活了两年——我并非由父亲抚养长大。他买了台来路不明的电脑，你知道的，总有人带些说不清来源的东西给他，可能是赃物。这台电脑引起了我的注意，我有个同父异母的哥哥，他只热衷修车。

That was pre United States. Well, I came, I spent two years with my father, I didn't grow up with my dad. And he bought this like computer, don't know, it's like, you know, people would bring him stuff that I don't know where it came from, it was stolen. And so he would buy stuff and he would just bring it home. And so the thing that caught my attention, I have a half brother and he just loved fixing cars.

后来父亲带回这台电脑，我哥哥看都不看。不知是电脑本身无趣，还是因为哥哥不用它刺激了我。我开始用WordPerfect和Lotus 1-2-3这些80年代的软件，那时我大概六七年级，在洛杉矶读初中。

And then my dad brought this computer and my half brother did not look at it. I don't know if it was a drone by it or drone the fact that my brother wasn't using it. And I just started using WordPerfect like Lotus one, two, three in the 80s. I was like in sixth, seventh grade. I did junior high in LA.

我沉迷于在那玩意儿上打字，折腾打印机。没有网络，没有交流，没有用户群。就我一个人在车库里摸索。后来回到尼加拉瓜时，我母亲竟有台笔记本电脑——80年代的尼加拉瓜！她是大学教授，兽医博士，带着笔记本电脑上班。我当时就想：你为啥会有笔记本？

And I just love typing on that thing and trying to get the printer to work. No internet, no communication, no user group. I just me in the garage, just trying to figure this thing out. And then when I went back to Nicaragua, my mother had a laptop I show up in Nicaragua in the '80s and my mom has a laptop, she's a professor, she's a PhD, a veterinarian at the university, and she has a laptop. And I'm just like, Why do you have a laptop?

那可是尼加拉瓜啊。于是我开始帮她处理技术问题，充当大学里的技术支持。其实多数时候是我把电脑弄坏，再送去IT部门修理。

This is Nicaragua. So I started helping her and being, like, her tech support at the university. Mostly, just broke the computer, and then I would take it to IT to get it fixed.

那么你大学毕业后的第一份工作是在IBM，对吗？

And then your first job kind of post university, was IBM. Is that right?

没错，是的。

Absolutely, yeah.

我们不一定非要讨论IBM，但请告诉我你遇见大卫·坎塞尔的那个时刻。

And we don't need to talk about IBM necessarily, but tell me about the moment that you met David Cancel.

是的，我在IBM与一群了不起的人共事。可以说1999年我在IBM的第一个项目是开发一个能查询人员电话号码的聊天机器人，叫做'蓝页'，类似于IBM内部的黄页。我学到了很多，也开发了很多，那是一段非凡的经历。但我最终还是受够了。

Yeah, I think I worked at IBM with an amazing group of people. I'll say that my first nineteen ninety nine, my first project at IBM I was building a chatbot that you could look up people's phone numbers. This is called blue pages, like instead of yellow pages at IBM. And I learned so much, I built so much, it was an amazing thing. But I just got fed up.

在40万人的公司里，你很难产生巨大影响。那里太政治化了。想象一下我在这个庞大体系中的处境。于是我不断尝试在波士顿科技圈建立人脉，这花了我很长时间。

400,000 people, you cannot make a huge impact in that company. It's just too political. Imagine me in a 400,000 person organization. And so I just kept trying to I kept making connections in Boston in the tech scene. It took me a long time.

后来我终于遇见大卫，这是我第一次在科技行业见到拉丁裔同行。我立刻就喜欢上他，觉得更愿意信任他加入他的初创公司——当时很多人邀请我加入他们的团队，但我不知道如何选择。选择他是因为相处起来很自在，我觉得他理解我，值得信赖。

Then I finally met David and first I'd never seen another HispanicLatino in tech, you know, I was like, Okay. Like, I liked him. I felt like I trusted him more to join his startup because a lot of people were asking me to join their startups and I did not know how to choose one. I chose him because I just felt comfortable. You know, I felt that he understood me and that I could trust him.

他做的第一件事是提议让我接手他在前公司的职位，我当时很诧异：'什么意思？直接去你前公司当工程副总裁？'但最终我决定加入他创办的Lookery公司。这是我第一次接触科技初创企业，因为我需要尝试IBM之外的可能性。

And the first thing he did is he offered me to take over his prior job at some other company, and I was like, What do you mean? Just be VP of engineering at your old company? But in the end, I really decided to join him in this company called Lookery. It was my first tech startup because I needed to try something outside of IBM.

是啊，从40万人缩减到40人左右是什么感觉？

Yeah, what was it like going from 400,000 people to 40 people or whatever

不，大概只有10个人。

the No, was like 10.

这就对了。

There you go.

我当时在洛厄尔市工作，卧室里摆了张办公桌，每天起床就坐下工作，24小时连轴转。那段日子就像同时照顾三个穿尿布的婴儿还全天候工作，精疲力竭。作为公司首席工程师，所有系统都得我亲手搭建——从IBM时期做无人问津的项目，突然变成全权负责的唯一开发者。

I was working, I had a desk and I lived in Lowell, Massachusetts. I had a desk in my bedroom, so I would get up and sit down and I would just work 20 fourseven. It was like that stage of my life was like three babies in diapers working 20 fourseven. It was so exhausting, like literally, I was like the main engineer in the company and I had to build everything. So I went from building stuff that nobody cared, nobody used at IBM to like I was the sole person responsible.

但那也是最棒的时光（抱歉我总说最棒）。正是那时我开始用AWS云服务，天啊！从以前在IBM要走采购流程订购刀片服务器，到如今按个按钮就能调用S3和EC2——作为光杆司令能独立构建整套系统而不求人，这种自由感简直震撼。这真是巨大的突破。

But it was the best I have so many best times, sorry. It was when I started using AWS. Ah! And I started using the cloud, and it was mind boggling from ordering and buying blade servers and requesting stuff through IBM, you know, procurement systems to do anything, to like pushing a button, S3 and EC2, and it was like I was just blown away that I could be this one person orchestrating and building everything without needing anybody. So that was a huge unlock.

后来你们创立了Performable，这是你创办的第一家公司，而Agency是第三家。Performable运营几年后并入HubSpot，我们就是在那时结识的。

And then you guys started Performable. So that was your first company that you founded. Agency is your third company that you founded. You guys started Performable and after a few years joined HubSpot, which is where we intersected. Yeah.

为听众说明下——相信Brian和Darmesh也会认同这个评价：早期HubSpot有精彩故事但产品平平。当你和David带着Performable团队加入后，原本只负责某个模块，最终却重构了整个产品体系。可以说HubSpot第二阶段'精彩故事+卓越产品'的双重优势，正是你们缔造的。

And for people's benefit, and I think Brian and Darmesh should agree with this characterization. HubSpot in the early days had an amazing story and, like, an okay product. And then you and David joined with your team from Performable. And, originally, the mandate was to build out a particular piece of the product, but then you ended up with responsibility for the whole thing and rebuilt it from scratch. And sort of the second phase of HubSpot, thanks to you guys, was great story, great product.

你能简单谈谈加入HubSpot的感受吗？以及随着时间的推移，你不仅如何随着组织规模扩大而成长，还在某些方面重塑了它？

Can you say a couple words about what it was like joining HubSpot and then sort of how over the time you were there, you you you managed not only to scale with the organization but to kind of reinvent it in some ways?

对我而言，这标志着职业生涯的重大突破。从摆弄电脑和WordPerfect，或是启动几个EC2实例、构建几个网络应用，突然要支撑5000名客户。我从服务Performable的20名客户瞬间扩展到5000名，这让我震惊。因为我更在意客户体验——以前Performable的客户会直接发短信说'出问题了'，我就打开电脑修改代码，然后让他们刷新页面。

To me, professionally, was a big breakout moment. Go from tinkering with computers and WordPerfect or launching a few EC2 instances and building a few web apps to really having to support 5,000 customers. I went from like 20 customers that performable to like 5,000 customers immediately. And that was a shocker to me because, one, what I care more about is customer experience, right? My customers are performable would text me and they would say, This is broken, and I would open my computer, edit the code and say, Reload.

但当客户量达到5000时，这种方式行不通了。我必须在各个方面学习规模化运营：我们面临巨大紧迫性，客户流失严重，产品完全无法使用，团队也运转失灵。所以我们必须重建团队、重构产品，并重新赢得客户信任——让他们相信这个产品真能兑现承诺。

When you have 5,000 customers, you can't do that. So I had to learn how to scale in every sense of the word, right? We had a huge urgency, we had major churn. The product did not work at all. And the team did not function either, so we had to rebuild the team, rebuild the product, and rebuild the trust with the customers, right?

同时我也要学会如何让自己作为领导者实现能力升级。其中最重要的经验就是规模化招聘。

That this product was going to actually achieve what it was going to do. And I also had to learn how to scale myself as a leader. And the most important thing I would say that I learned there was like hiring at scale.

我正想请教这点——记得你在创办Klaviyo之前就聘用了Andrew Bilecki，应该还招募过Christopher O'Donnell

I was going to ask you about that actually because you I believe you hired Andrew Bilecki before you started Klaviyo. I I believe you hired Christopher O'Donnell

没错。

Yeah.

在他后来开发HubSpot CRM并自立门户之前。你确实招揽了几位非常杰出的人才。要知道，波士顿一家处于发展中阶段的营销软件公司能吸引到这样顶尖的人物并不常见。

You know, before he went on to build HubSpot CRM and and start a company of his own. I believe you had a handful of really exceptional hires. Yeah. Yeah. And it's not obvious that a mid stage marketing software company in Boston would be able to attract such great people.

我很好奇，你招聘时看重应聘者的哪些特质？之后又是如何帮助他们在职业道路上取得如此大成长的？

So I'm curious, what were you looking for in people that you hired? And then maybe what did you do to help them have such growth in their careers? I

我认为许多人都非常优秀，关键在于既要给予他们机会，也要以身作则展示成为杰出者意味着什么。可以说，我的成功既源于早期的自我特质，也离不开一路上许多人塑造了我对可能性的认知——比如布莱恩和达梅什——他们让我渴望追求并达成同样的成就。像惠特妮、现任HubSpot执行副总裁的贾里德等人至今仍在运营整个体系。

think so many people are exceptional and it's a combination of giving them a chance and also role modeling after them what becoming an outlier really means. And so I would say, like, am I as successful as I am because of who I was so early, or is it because a lot of people also shaped me along the way to teach me what was possible, And made me want to go after and accomplish the same things, just like Brian and Darmesh, for example, right? And so I think all these people, there's so many people, there's Whitney, there's Jared, current EVP at HubSpot, that is still there running the entire system.

我都忘了惠特妮，确实如此。

I forgot about Whitney, yeah, good point.

还有更多人呢？这很奇妙。倒不是我刻意记录，但能影响他人生活和职业发展确实让我倍感自豪。其实他们当时处境和我一样——我在车库里捣鼓WordPerfect软件，他们做着你能想到的最底层工作。我的做法是给人机会，通过交谈测试其才智，看能否跟上节奏，是否有渴望和韧劲，然后带他们上路。之后的路由他们自己选择。

Yeah, there's a few more, right? It's amazing. Not that I'm like keeping track of them, but it just makes me really proud, like, to have people that have been able to influence their lives and their careers, right? But it was really they were at the places where I was. I was in a garage just tinkering with WordPerfect.

我不看重学历背景，不关心你从哪来或自称做过什么。说实话，如果你只有这些，我反而会立即排除考虑。

They were doing, you know, the crappiest jobs that you could possibly think of. And so my thing is I give people a shot, right? And I can test their intelligence from the conversation, see if they can keep up, see if they have hunger, they have grit, and then just bringing them along for the ride. And then they get to choose the path that they take after, right? And so I think that I just I don't hire for credentials.

有点可疑。

I don't hire for where you came from, what you said you did. I actually, like, if you have all that stuff, I'm kind of like rule you out immediately.

确实可疑。

Little suspicious.

-太可疑了。完全可疑。

-It's sus. Totally sus.

没错。我们简单聊聊Drift吧，因为这是你的第一个创业项目Performable之后的第二个项目，又是你和David Cancel一起做的。我知道你在Drift的经历对你现在正在创建的agency项目有很大启发。你能简单说说Drift是什么，以及Drift的哪些方面让你产生了现在做agency的灵感吗？

Yeah. Let's talk about Drift real quick, because so startup number one, Performable. Startup number two, Drift, which again was you and David Cancel. And I know that your experience at Drift helped to shape the idea for agency. So can you just say a word about what was Drift and what aspects of Drift gave you the insight for agency, which you're building now?

我觉得在HubSpot学到最重要的是价值实现时间这个概念，就是客户需要多久才能体验到你们为他们打造的产品价值？在HubSpot，这个过程很长。你得建博客、做SEO排名、获取潜在客户、建立品牌。所以从第一天起你就在引导客户：我们这是一场长期旅程。而且是以低价实现的，HubSpot能达到今天的营收规模真是个奇迹。

I mean, I think that the thing that we learned at HubSpot was time to value, right? It's like how long does it take for the customer to experience the value and the product that you're building for them? At HubSpot, takes a long time. You have to build blogs, you have to build ranking in SEO and Google and you have to get leads and you have to become a brand. So you're coaching the customer from day one to like, know, like we're in here for a journey And at a low price, which is an unbelievable achievement for HubSpot to be at the revenue that it is today, right, at the scale.

但在Drift时，我和David一直在思考：怎样才能缩短价值实现时间？所以我们初期很有耐心，花了大约两年时间在沙漠里徘徊，试图找到方向...

But I would say that at Drift, we were kind of like, it was almost something I always kept thinking with David and we were like, How can we get time to value? And so we were patient in the beginning, we spent about two years wandering through the desert trying to figure out what should we work on that

对——记得最初是做HR软件的。

was -Yeah, remember it began as HR software.

-天啊别提了。这真是个糟糕的玩笑。我们当时想做Glassdoor那种，是个匿名秘密应用。

-Oh, God. Yeah, let's not talk about that. God, there's a bad joke in there. We were trying to do, like, Glassdoor. It was like a secret app at that time.

记得吗？员工可以匿名吐槽公司内幕。我们在HubSpot时就痴迷团队动态和文化，但后来不断转型。每个月都在转。每次Pat来波士顿看我们都会问：你们现在做什么项目？哦，新的Dropbox。

Remember, People could say secretly, anonymously, what's going on at their company. We obsessed with team dynamics and culture at HubSpot, but then we kept pivoting and pivoting. We would pivot every month. Every time Pat would stop by in Boston to see us, So what are guys working on? Oh, a new Dropbox.

就像，你们现在在做什么？一个新的Pinterest。现在呢？SurveyMonkey。基本上每个月我都会告诉团队，那个主意并不好。

It's like, what are doing now? A new Pinterest. What about now? SurveyMonkey. It's like literally every month I would just tell the team, that was not a good idea.

我们得尝试另一个方向。于是，我们花了两年时间保持极大耐心，寻找真正想要的东西。而我意识到的是，你必须回归熟悉的领域。经验会累积，你需要坚持。所以我们经历了一次大偏离，最终又逐渐回归到应用、服务、客户的核心。

We're to try another one. So one, we spent two years being super patient and finding something we want. And the thing that I realized is like, you got to go back to what you know. Experience compounds and you got to just like stick with it. And so we went on a huge diversion and then came back to, like, closer and closer and closer to the app, to service, to customer.

我们当时就在HubSpot隔壁，所以不会涉足营销领域。我也不想再造一个CRM系统。于是我们从客服支持切入，然后转向应用内功能，后来又回到潜在客户开发。核心理念是：如果部署这个方案，就能立即自动获取更多销售线索，无需人工干预。

And we were like, were next door to HubSpot, so we were not going to do marketing. And I was not going to build another CRM. And it was like, no. And so we started focusing from support, then we went in app, and then we went back to leads. And the idea was to, if you could deploy this, that could immediately start getting you more leads without any human effort.

这就是规模化关键点，对吧？还有自动化。我们设计了一个聊天机器人充当前台接待，在官网欢迎访客。它不会要求填写表格，而是主动提问、提供价值信息、安排会议、转接对应销售，并直接发送通知到销售手机——快速建立联系，消除摩擦。这很棒，不是吗？

That was the scaling point, right? And automation. So we had a chatbot that would be the receptionist, the welcome reception on your homepage. Instead of asking you to fill out a form, could ask you questions, it could inform you delivery some value, book a calendar meeting, route it to the right rep and send a notification to the cell phone of the rep and just make connections faster and remove the friction. And that was fantastic, right?

这个方案确实有效。我认为它帮助我们实现了起飞。我们做对了很多事。但现实是，这个产品深度不足——因为它只解决了客户旅程中最基础的一环，仅仅完成了初次连接。

That worked, right? I think that that helped us take off. We did many things right. But the reality, this product was shallow, right, in that it wasn't too deeply penetrated because it was only solving one basic step in the customer journey. It only did like the first connection.

但随后出现的问题是：所有人都会问'为什么我再次访问网站时系统不认识我？为什么不能延续上次对话？我已经解释过这些了。为什么在应用内使用时，系统不知道我之前的支持工单和沟通记录？'于是我萌生了一个构想：如何能全面掌握客户的所有信息？

But the problem that arose then after is like everybody would be like, Well, I came back to the website, why doesn't it recognize me? Why doesn't it pick up where it left off? I already had this conversation, I already explained this. How come when I'm inside the app it doesn't know, you know, what I said before, my support tickets and my whatever, right? And so there was this dream I had of like, how could we always know everything about the customer?

以及如何在我们每次交互中运用这些信息。我们应该了解自己的客户，对吧？但痛苦在于当时缺乏实现这个构想的技术能力。此外另一个困境是：我们快速构建了品牌和产品，拥有惊人的上市速度，却缺乏规模化运营的能力。

And how could we always apply that information in every exchange that we have. We should know our customer, right? And so it was a pain that we didn't have the technology to be able to deliver that. And so I think that was one struggle. The other struggle I would say was that we built quickly, we built brand, we built product, we have amazing speed to market, but then we didn't have the ability to scale.

我们这里总是认为，只要雇佣更多人、筹集更多资金，不必担心开支，问题自然会解决。以后总能扩大规模。但我认为这种想法从根本上就是错误的。说实话，在这个AI时代，我不知道人们是否还明白这一点。

Here we always think, well, just hire more bodies and just raise more money, don't worry about spending. It will resolve itself out. We can always scale it later. And I think that that's fundamentally flawed. I don't know if people still get that, to tell you the truth, in this age of AI.

我认为这正是AI存在的意义——证明这种观点的谬误。但其他人却说，先招人再说，以后再操心。在Drift的经历让我学到最重要的一课是：我始终热爱与客户交流。我喜欢他们发短信或打电话给我。作为外向者，我从中获得能量，这种人际关系让我充满活力。

I think that is the whole point of AI is to prove that out. But other people are like, Just hire and we'll worry about it later. And I think that that was my biggest lesson at Drift that I always liked talking to my customers. I always liked to text me, call me. I'm an extrovert and I just thrive on that, on relationships.

当Drift的客户数量达到成千上万时，我再也无法与他们一一交谈，无法了解每个人。通常当客户经理联系你时，往往是因为某个可能只通过一次邮件的客户即将流失——这种时候通常已经无法挽回了。即便我承诺给予全世界、承诺随时支持，成功率可能只有20%，这很肤浅、很虚伪，根本不是事实，你明白吗？

And when we got to the point when we started having thousands and thousands of customers at Drift, I just could no longer talk to them. I could no longer know them. Usually when a CSM calls you, because it's a churn of somebody that I might have emailed once, commonly hasn't saved this churn, it's just too late. You can't do that. I mean, I might have a success rate of like 20% if I promise them the world and I'm going to be there for you, that's shallow, that's hypocritical, that's not the truth, you know?

我深切体会过规模化带来的痛苦：当团队膨胀到600人、800人时（我一直记着人数），你把更多时间花在管理员工、他们的晋升、成长、入职、招聘、团建和全员大会上。这些事务吞噬了我们，让我们忽视了最核心的客户和产品。正是在这里我们付出了代价，未能实现最初梦想——打造一家传奇企业。

And so I felt that pain intimately of what happens when you scale and what happens when you have 800 people, 600, I keep track, you know, and you spend more time managing the people, their promotions, their growth, their onboarding, their recruiting, their entertainment, the town halls. And I think that just consumed us and realized that took our eyes from the ball of the customer and the product. I think that that's where we paid the price and not being able to build a legendary company as we dreamed to build from the beginning.

而Drift最终估值仍超过10亿美元。你的第一个创业项目加入了HubSpot，第二个创业项目规模更大。那时你简直就是美国梦的活标本——从赤手空拳来到这个国家，凭借智慧、拼搏和惊人意志力，创造了这些非凡成就，彻底改变了你和家人的生活轨迹。

And Drift was ultimately still worth more than $1,000,000,000 And so your first startup success joined HubSpot. Second startup success of greater magnitude. And at that point, you're you're kind of like a walking example of the American dream. Know? Like, you came to this country with nothing and through smarts and hustle and sheer force of will, you know, you built these amazing things and put your life in a totally different position for, you know, for you and your family.

你本可以功成身退，就此结束职业生涯。为什么选择重头再来？

You could have ridden off into the sunset and just called it a career. Why do it all over again?

其实我确实隐退过。大约七个月时间里，我像其他人那样周游世界。但说实话，这种生活异常空虚无聊。当时我正在巴西塔塔朱巴、杰里科阿科达和朋友玩风筝冲浪。

Well, I did. I did ride into the sunset. I did that for, like, seven months. I think that people traveled the world, did all this stuff, but it's extremely empty and boring, I would say. But I was kite surfing in Tatajuba, Jerichoacoda in Brazil, you know, with some friends.

壮丽的地方。这是我第一次真正地断开连接。就像，这是我离开Drift后的第一次旅行。我心想，哇，我没有会议，没有任何事情要做。整个早上，我们就这么躺在小屋前，望着这片海洋，空无一人的海滩。

Majestic place. This is the first time I'm actually disconnecting. Like, this is my first trip after I leave Drift. And I'm like, Wow, I have no meetings, I have nothing to do. And I'm just like, the whole morning, we're just staring in this cabana, just looking at this ocean, empty beaches.

这是巴西东北部一个非常偏远的地区。我不禁感叹，哇，我需要学习如何冥想，如何与自然融为一体。下午我们会去玩风筝冲浪。然后ChatGPT发布了，你知道的。我就这样向每个人展示并解释着。

It's a very remote area in the Northeast Of Brazil. And I'm like, Wow, I need to learn how to meditate and how to just be one with nature. And in the afternoon we would go kitesurfing. And then ChatGPT launches, you know. And I'm just like showing everybody and explaining.

我甚至不知道该如何解释。就像，它怎么会知道如何回答这个？所以，当我第一天真正精神上放松下来时，恰逢ChatGPT发布，我心想，这是人类最伟大的时刻。然后我就想，妈的，我得重新投入工作。你知道吗，感觉就像我为此奋斗了一生。

I don't even know how to explain it. It's like, How does he know how to answer this? And so the first day that I'm off really mentally is the day Chad GPT launches, and I'm just like, This is the greatest moment in mankind. And I'm just like, Fuck, I got to go back in. You know, it's like, I feel like I worked all my life.

我一直认为自己是个冒牌货。但现在不是了。你知道，我们做到了。但问题是，我这一生都在想，我没上过那所学校，没在硅谷打拼，我在波士顿，这家公司倒了，全都是这些事情。而且因为这样，我没有足够的资源，没有人脉，不认识投资人，不懂这个，不会英语。

I always considered myself an impostor. Not anymore. You know, we achieved that. But the thing is, like, all my life I'm like, I didn't go to this school, I didn't build in the Valley, I'm in Boston, this company fell, it's only all this stuff. And like, I didn't have enough because of this, I didn't have this network, I didn't know investors, I didn't know this, I didn't know English.

而现在我想，我他妈什么都懂，认识所有人。我建立的系统足以向数百万用户证明，还有收入、销售等等。当那些什么都不懂的人都在搞AI创业时，我怎么能坐在这里？门都没有。

And now I'm like, I fucking know everything. I know everybody. I built the systems enough to prove to millions of users and this and that and revenue and sales. And I'm just to sit here when all these people that don't know anything are like building AI startups? Fuck no.

我要重新投入，你知道吗？我觉得这就是那个时刻。虽然说起来可能有点俗套尴尬，但作为一个拉丁裔，我感到一种责任。要知道，大多数人做梦都想处于我现在的位置。如果浪费这个机会，回到当年在尼加拉瓜家里没饭吃的时候——浪费资源、机遇，还有那些用时间和努力换来的东西，就只是在巴西玩玩风筝冲浪，或者在斐济冲浪。

I'm jumping back in, you know? I was like, this is the moment. And I will say that the cheesy, cringey answer to this is also, as a Latino I feel an obligation. Know, it's like most people would die to be in the position that I am in right now. And to squander, going back to that day in my house in Nicaragua with no food, To squander resources and opportunity and just so, like, time earned, work earned, you know, to just sit around and just kite surf in Brazil, you know, or like surf in Fiji.

我觉得这简直是浪费。只要我还活着，就必须做点什么。说到成为榜样，人们想看看和你相似的人能取得什么成就。我想，这次我不能袖手旁观，对吧？所以那时我打电话给你说，嘿，我能来大本营吗？

I just think it's a waste. While I'm on this earth, I have to do something. And going back to being a role model, people want to see what someone that looks like you is capable of achieving. And I'm like, I can't sit this one out, right? And so that's when I called you and I'm like, Hey, can I come to base camp?

我能重新连接上，被这些东西点燃吗？我觉得我来到了AI的上升期，那一刻就像在说，咱们他妈干吧。

Can I get reconnected, fired up with this stuff? I think I came to AI ascent and that's when the moment it was like, Let's fucking go.

-是啊。-你懂吧？-对。2023年初，我记得你参加了AI Ascent，然后开始和OpenAI接触。-是的。

-Yeah. -You know? -Yeah. Early twenty twenty three, I remember you came to AI Ascent, and then you started talking with OpenAI. -Yes.

-他们当时遇到的一个问题是，许多企业第一反应都是：能不能把这东西部署到面向客户的场景里，某种程度上有点像Drift的业务模式。我记得你就此重新介入，后来把这个想法发展成了现在的Agency形态，已经大不相同了。能简单说说吗？我们之前聊邮件时提到过，但能不能用几句话解释下，什么是Agency？它为客户解决了什么问题？

-And one of the problems they were having was they had all these enterprises whose immediate thought was, can we deploy this thing in a customer facing way, kind of like the Drift business in some ways. And so I remember you kind of jumped back in on that and then evolved the idea into what agency has become, which is very different. And can you just say a word we kind of backed into this earlier, talking about the email stuff. But can you just say a word like, what is agency? And what problem does it solve for customers?

为什么这个问题很重要？能简要谈谈吗？

Why is that problem important? Can you just say a few words about that?

没错。全世界所有人面临的共同难题是：他们难以提供优质的客户体验。而目前唯一的解决方案不是靠技术，而是依赖人力。所以由人类主导的客户体验交付，几乎是件不可能完成的任务。Agency正在构建的，正是由AI主导的客户体验。

Yeah. The problem that everyone in the world has is that they struggle to deliver a great customer experience. And the only way that they can solve that problem today is not with technology but is with humans. And so, human led, deliver customer experience, it's almost an impossible thing. And so what Agency it is building is AI led customer experience.

如何真正突破想象极限地扩展客户体验，而不需要在每个环节都安排人力？懂我意思吗？这就是我要解决的核心问题。这个想法逐渐成型——我就是热爱生活，对吧？现在年纪大了，因为你能追溯所有经历，所以我刚给你讲了不少片段。

How do we truly scale customer experience beyond our wildest imagination without needing a human in every step of the way? Right? So that's what I'm really solving for. And it kind of like came through I just I love life, right? And now that I'm older because you can trace back all the steps and so I gave you a whole bunch of them.

但巴西、ChatGPT、AI Ascent，你引荐我认识OpenAI。OpenAI说他们只有200人，不知道如何应对这么多客户需求，也没法教所有人写提示词。他们问：你愿意帮忙吗？

But Brazil, Chad GPT, AI Ascent, you introduced me to OpenAI. OpenAI says we're just like 200 people. We do not know how to talk to all these customers. We can't tell them how to prompt stuff. Would you help us?

当时有一批公司认为，'让我们从AI服务开始做起'。于是我也这么做了，成立了一家咨询公司，开始为各种客户提供服务，比如NBA、红牛这些大公司。其中一个是Klaviyo，你知道的，安德鲁。我问他需要什么帮助，他说需要提升客户体验。

There was jump there's a cohort of companies that were like, Let's build services in AI and let's start there. And so I kind of did that, built a consultancy and just started servicing all kinds of customers, like the NBA, Red Bull, crazy companies. One of them was Klaviyo, you know, Andrew. I said, What do you need help with? And he's like, I need help delivering customer experience.

就像，如何帮助客户理解使用Klaviyo的潜力？我们该如何提供建议？这本来就是客户成功经理（CSM）的工作，对吧？那么问题来了，如何让每个CSM都成为专家？让每个招聘的CSM都成为Clearview或电商营销领域的顶尖专家。

Know, it's like, How do I help my customers understand what the potential is of using Klaviyo? How do we give them advice? Which is the job of a CSM, right? And so, it's like, how can a CSM know be an expert? Every CSM that you hire be an expert, the number one in the world expert in Clearview itself or in e commerce marketing.

这就是AI的用武之地。当他最初提出这个要求时，我直接说'绝对不可能'。我甚至不确定能否做到——想象一下，如果真实现了，我的公司规模可能超过他的。结果他说'只管去做'，就这么简单粗暴。

So that's where AI comes in. And when he first asked me to do this I said No fucking way. I don't know if that could if I could do that, imagine, like, would my company would be bigger than yours. He's like, Just do it. So he just, you know, rolls reverse, right?

他坚持说'去实现它'。于是我们构建了一个模型，能理解每位客户的所有互动，分析哪些策略有效、哪些无效，能对整个公司所有用户的使用情况进行分类，并开始提供定制化建议。突然间我豁然开朗：如果我能自动生成季度业务报告呢？如果能制作讲解视频为客户节省时间呢？

And he's like, Go do it. Go do things. And so we built a model that understood every customer and every one of their interactions and understand what was working and what was not working and be able to classify the entire usage of every user in the company and start to be able to deliver tailored advice, right? And then the whole thing started clicking for me where I was like, what if I can just deliver the QBR? What if I can just like do you know, create a video explaining that and saving time to the customer?

那一刻我意识到：这就是我想解决的痛点。虽然当时还不确定能否成功，但就在同一天，布莱恩·哈利根发信息给我——之前他一直在'追求'我合作——那天他说'有个创意给你，我们来颠覆世界吧'。

And so then I was like, this is great. Like, this is the moment where like, this is what I want to solve. And I wasn't sure, right, that that was going to be it. Literally that day, like, Brian Halligan sends me a text, like, you know, had been sending me all these messages, romancing me, then that day he's like, I have an idea for you. You know, it's like, Let's disrupt the world.

他说要颠覆整个行业，我反问'你在说什么？核心是人的问题。我们无法规模化，怎么解决客户体验这个难题？'

You know, let's disrupt this entire industry. And I'm like, What are you talking about? It's about the people. We can't scale. How do we solve this problem in customer experience?

但最终我说'干吧'。后来我们联系了你，就这样创立了代理公司。一切开始飞速发展，明白吗？

Let's do it. And then we called you and that's how agency was born. Things just accelerate, you know?

从IBM的聊天机器人到现在聊天功能大行其道，从在Drift只解决交易性对话到现在想要管理整个客户体验，这真是个完美的轮回时刻。简直不可思议，完整的循环。

It must have been a really cool full circle moment going from chatbots at IBM to now chat is working and going from you're only solving the transactional part of the conversation at Drift to now you want to manage the entire customer experience. It's incredible, full circle moment.

是啊，我觉得很多人都在挣扎。他们正经历着关于AI的身份危机，比如：我该怎么办？我会被取代吗？在我看来，我发现很多创始人或普通人都在纠结：我的目标是什么？我该做什么？

Yeah, I think that a lot of people are struggling. They're having an identity crisis with AI of like, What should I do? Will I be replaced? And to me, I think what I've learned, what I see a lot of founders struggle with is, or people, is like, What's my purpose? What am I going to do?

而且我要做什么才能持久？所以我要做什么大事？因此对我来说，发现客户体验这个问题是如此令人宽慰和安心。这是最困难的事。我不是说它容易，但我实在想不出这个问题最终要怎么解决。

And what am I going to do that's going to last? So what am I going to do that is big? And so to me, finding the problem of customer experience is so soothing and reassuring. It's like the hardest thing. I'm not saying it's easy, but I just can't think of how this will ever be solved.

就像无论什么样的客户体验——可以说世界上大多数客户体验都很糟糕。我能想到最好的大概就是豪华酒店了，但有时候...（虽然我不想点名我住的酒店）可当我让他们给我送点药品什么的，他们就说'托雷斯先生，这要收4美元'。我花了上千美元住这里，送个泰诺还要收我4美元？

It's like whatever customer experience we could like most customer experience in the world sucks. I would say the best I could think of maybe could be like a fancy hotel, you know, like sometimes, but I don't want to name my hotel that I stay here. But now if I ask them to send me some delivery, a medicine or something, they're like, That'll be $4 Mr. Torres. Like, why am I paying thousands of dollars in this room and you're charging me $4 for delivering me, like, my Tylenol?

还有我想要的食物，我不想下楼取。结果他们说'这要收7美元'。这大概就是当今能享受到的最顶级客户体验了，所谓被 pamper 的待遇不过如此。

It's like, and then I got this food that you want, I don't want to come down. I was like, that'll be $7 And and this is like the greatest experience you can get today, I think, in a customer. Like that's where you get pampered.

嗯。你们现在有代理客户吗？有什么特别喜欢的客户故事吗？

Yeah. Do you have any customers on agency today? And any favorite customer stories?

红杉资本经典三连问：营收多少？增长...

Classic Sequoia is like, What's the revenue? What's the

曲线？我们来谈谈ARR吧。

curve? Let's talk about ARR.

我们来谈谈ARR。

Let's talk about ARR.

我们查看了你们的毛利润

We went through your What's gross

利润率？是啊，因为你们想象一下。其实我们可以讨论这个。这简直一团糟，我不知道我们怎么承受得起这些LLM成本。

margin? Yeah, because you guys imagine. Actually, we can talk about that. This is a shit show. I don't know how we're gonna survive with this LLM cost.

我觉得这整个就是个金字塔骗局。

I think it's an entire pyramid scheme here.

嗯，其实，关于这点说两句。

Well, actually, say a word about that.

对，对。我是说，我觉得我们有超过50个客户在使用这个产品对吧？我才不在乎那些数字什么的。

Yeah. Yeah. I mean, I I think so we have over 50 customers using the product. Right? I don't give a shit about numbers and stuff like that.

我在为他们构建产品，解决问题，感到由衷欣喜。原本我想获取更多客户，但布莱恩说，专注于现有客户，让他们取得惊人成功。我当时就想，

I'm building for them, I'm solving, I'm really delighted. I was trying to get more customers, and Brian said, Stick with the customers you have. Make them so wildly successful. And I was like,

是的，我认为关键在于虽然可以开发更多客户，但你现在应该集中精力让这50个客户对产品极度满意。

yeah, I think that's important is that there are more customers that you could go get, but you're focused right now on just making these 50 deliriously happy with the product.

这正是我的现状。回到成本问题——我们一直在与他们合作，这个过程很艰难...因为过去我会在浏览器加载出雏形时就急着收费，哪怕只要一美元。而现在我完全改变了想法，

That's exactly where I am, right? And so back to the cost stuff. So we've been working with them and I have a really It's hard going multiple times around because this time around, in the past, I would be like, the minute I had something that kind of like loaded on your browser, like I wanted to charge you. You know, it's like, a dollar. I was like,

明白，

yes,

这次我不要你们的钱。我只想确认你们对产品着迷，它足够有粘性，能渗透到你们公司的每个环节。太多创始人误以为做出产品就万事大吉，这简直是妄想。

I got a dollar. This time I'm like, I don't want your money. It's like, I want to know that you're obsessed and that it's so sticky and that I'm going to do everything at your company. And I don't want to make my There's so many founders that they think that just because they built something that they got it, right? People are delusional.

有些人年收入五千万美元，却仍妄想产品没问题。这次我真心想做好——打造让人主动砸钱、惊呼'这产品独一无二'的东西。这就是我现在的征程，要创造真正有价值的事物。

People have $50,000,000 in revenue and they're delusional that that product works. You know, this time around, I really want to do it right. I want to build something that people are throwing money at me and saying like, This is really, there's nothing like it. This is my journey now, right? So I want to build something real, right?

不像大多数SaaS产品——包括我过去做的——只会给用户增加工作量。我要做减法工具。所以我现在更有耐心，也许这会让我吃亏，也许不会。幸好有理解我的投资人，他不催我。但真正困难的是我现在试图弄明白...

That is not going to like just I think most of SaaS out there that we ever built, that I ever built, just created more work for the user. I want something that takes the work off, right? And so I'm being more patient, right? And maybe it hurts me, maybe it doesn't, but I got great investors like that, and he's not rushing me, so sometimes. But I would say that what's really hard is I'm trying to understand.

我的首要目标，就是创建一家年收入达十亿美元但员工不足百人的公司，明白吗？我是最早提出这个想法的人之一，当时人们都觉得我疯了，真的疯了。但我是认真的。我的团队大部分是工程师，我们专注于产品开发，因为大部分资金和人力都投入到了市场推广和客户体验上。

My number one goal, right, is I want to create a company that is a billion in revenue with less than 100 employees, right? I was one of the first ones saying out there, and people were like, You're crazy, you're crazy. And I really mean it, right? Most of my team is engineers, right? And we're just building the product because where most of the money and where the headcount goes is in GTM and customer experience.

因此我想解决这个问题，确保我们能为未来打造公司。尽管我们身处世界上最具创造力的国家，但我看到产品领域严重缺乏创新——所有人都在开发相同的聊天工具、会议记录器、笔记摘要和邮件草拟功能，对吧？

And so I want to solve that, I want to get that right so we can build the companies of the future. There's a lot of lack of creativity and we're in the most creative country in the world, but what I see out there in products is everybody's building the same chat, meeting recorder, the note summarizer and the email drafter, right?

我记得你创办Agency时，HubSpot有句原则叫'他们往左，我们就往右'。当时AI销售开发代表（AISDR）公司大行其道，而你的逆向选择是专注业务后端，因为那里才是真正的痛点所在，也是价值创造的地方。你在众人趋同时选择了逆向而行。

Well, I remember when you started Agency, one of the HubSpot principles or norms has always been when they zig, we zag, right? And I think when you started Agency, the zig was all these AISDR companies. And your zag was, no, no, we'll let them take the front end of the business. We're going to deal with the back end of the business because that's where the pain actually is, that's where the value is built. And so I think you of zagged when people were zigging.

过去一年证明这个决策很正确。不过能否谈谈当前AI的能力边界？在你开发的产品中，AI哪些方面表现出色？哪些方面还存在...

And I think over the last year or so, that's proven to be a good decision. But can you say a word about what can AI do today? What can it not do today? And so in the product that you've built, where is AI or the product itself really nailing it? And where do

你觉得可能还存在能力缺口？如果乐观地说，理论上AI的能力是无限的...

you feel like there is maybe still a capability gap in terms of what can be done? Well, think that if we're optimists, AI is infinitely capable in the theoretical

假以时日的话。

In the fullness of time.

没错，假以时日。但目前的AI其实并不那么聪明。你需要写大段文字才可能得到正确答案'42'。如果输错单词或重试，答案就完全不同。这部分正是当前人类大脑难以适应的。

In the fullness of time, right? But today, it's actually not that smart. And so, you have to write paragraphs and paragraphs to maybe get the right answer, 42. And so, you know, if you type the wrong word or if you try it again, it's just a different answer. And so that's the part that today's human brains just can't cope with it.

因此，当你试图为客户提供体验时，你正在努力弄清楚你到底想要做什么。你是只想找个更好的词回复客户，还是想让AI真正处理并管理客户体验？我认为我们目前的困境在于，如果我们甚至不能信任AI来起草你自己即将点击发送的邮件，那么让AI每天自主为上万名客户做决定的情况会有多糟糕？所以我们构建了代理的第一部分，这也是大多数人现在正在尝试做的，即所谓的‘客户360度’基础方案——将所有信息集中到一个地方并能够提问。但大多数人没有意识到的是，如果你需要人工在聊天框里输入内容，那从一开始就注定失败。

And so when you're trying to deliver experience to the customer, you're trying to figure out what is it that you're exactly trying to do. Are you trying to just come up with a better word to reply to the customer or are you trying to let the AI actually do and manage the customer experience? And so I think where we're struggling is that if we cannot even trust the AI to draft your own email that you're about to hit the button, how much worse is going be for us to let the AI make decisions on 10,000 customers on a daily basis on its own? And so we built the first part of agency, which is what most people are trying to do today, which is this, you know, Customer three sixty, foundational, let's bring all that information into one place and be able to ask questions. But what most people are not realizing is that if you require a human to go type into that chat box, you're dead on arrival.

所有人都很困惑，即使在硅谷，人们也在开发那些只是帮助人类完成某些任务的工具。我已经受够了这些。我不想再做这种事。在ChatGPT里打字让我无聊得要死。我现在是CEO了，被惯坏了，正过着最好的生活。

Everybody's so confused and even in the valley of like building tools that are like just helping a human do something. I'm done with that. I don't want to do that. It bores me to death to type stuff into ChadGPT. I'm a CEO now, I'm spoiled, I'm living my best life.

我只想告诉人们一次要做什么，然后他们就能执行一万次。这就是我真正希望AI为我做到的。我们该如何指导AI无限次执行一个任务？大多数人却卡在这里。别提‘代理’这个词，否则我会停不下来。

I just want to tell people what to do one time and they do it 10,000 times. And so that's kind of what I really want you know, AI to do for me. It's like, how can we instruct AI to do one task infinitely? And most people are stuck. Like people don't get me started on the word agents, right?

每个人都在谈论代理，比如‘会议结束后给我邮箱发个摘要’。谁在乎这个？我随时可以问这个问题，但关键是我能用它做什么？所以我正在与合作的公司一起回归最基本原则，因为人们总问我：‘能生成客户仪表盘吗？’然后呢？

Everybody talks about agents and like, okay, send me a summary after my meeting to my email. Who gives a shit? I can always go ask that question but it's what do I do with it? So what I'm doing is I'm doing the very first principles in the companies that I'm working with because people just keep asking me, Oh, can you generate me a dashboard about the customers and then so what?

是啊，你想

Yeah, what do you want

用这个仪表盘做什么？

to do with the dashboard?

你想

What do

对着仪表盘操作时说，哦，我想点这个然后做那个。为什么我不能这么做？接着你要我做什么？让我去跟进续约？还是让我去告诉销售代表？

do with the dashboard and say, Oh, well I want to click on this and then do this. Why can't I do that? And then what do you want me to do that? You want me to go after the renewals? You want me to go tell the rep?

然后你要对销售代表说什么？哦，我要让他们发邮件给客户。为什么我不能直接发邮件给客户？对吧？这就是我现在乐在其中的探索过程，但确实很艰难。

And then what are going to tell the rep? Oh, I'm going tell them to email the customer. Why can't I just email the customer? Right? And so that's the journey I'm in that I'm enjoying, but it's really hard.

我得给整个商业世界来次认知重置，因为他们早已习惯这种运作方式。记得布莱恩生日那天，我带着眼镜去他家，他震惊得说不出话——他从未见过那副眼镜。他一位老友（布莱恩年纪大了，我也老了，现在说话可以肆无忌惮）

I have to deprogram the entire business world because they're used to working and operating this way. You know, Brian had his birthday. I was at his house, showed up with his glasses and then he was like in shock. He's never seen them before. And so then it was one of his old friends, because Brian is old, I'm old, I can say anything I want now, I'm old too.

他那位退休的风投老友像先知般对我说：埃利亚斯，我的建议是——忘掉你学过的所有东西。忘记模式识别。抛开你的经验。统统放下。

One of his old retired VCs he came to me like an oracle. And he was saying, Elias, my advice to you: forget everything you've ever learned. Forget the pattern recognition. Forget your experience. Forget all that.

那些都是狗屁。要勇于尝试新事物。他劝我放手，不要依赖过去。因为像我们这样的资深创业者容易陷入固有思维。这就像一场认知革命，对吧？

That's bullshit. Just got to like try new things. And he's like telling me to let go and not like rely on the past. Because sometimes like experienced founders, we get stuck in that way. And it's like this deprogramming, right?

要勇往直前拥抱未来。我的挑战在于：如何成为客户的变革推动者，让他们真正相信并停止纠结旧方法，专注于能与顶级客户直接沟通的正确方式。其实我们很多传统做法根本没必要，只是受限于技术。但要让AI成功执行'阅读仪表盘-生成方案-执行'这样的流程，光实现其中一环的稳定性就需要大量调试。不过内部可变革的空间其实很大。

To just charge into the future. And that's my challenge of like how do I become a change agent to our customers, to really believe and stop worrying about the things that we were doing before and focus on the right thing that they can do and they can have that one on one with their customer, with their top customer. But a lot of stuff that we're doing is really unnecessary because we just didn't have the technology. But it's going to take a long time for AI to be prompted into success of like, Read this dashboard, make this concrete and do this Just even doing one of those workflows consistently every time, it's going to take a lot of prompting. But there's a lot of stuff that you can bring change internally.

作为客户体验平台，我们能解决许多内部效率问题，但最有价值的还是直接将成果交付给客户。最后一个问题。

As a customer experience platform, we can bring things to resolve a lot of internal inefficiencies, but the most valuable thing is delivering that directly to the customer. Last question.

让我们以积极的态度结束。你对AI未来最乐观的期待是什么？

Let's end on a high note. What are you most optimistic for in our AI future?

我认为我最乐观的是希望人们能更多地发挥他们的主观能动性。你看，我觉得我们整个历史都在拼命工作。我认为我们需要停止如此专注于艰苦、艰苦、艰苦的工作，而是能够做我们最擅长的事情，那就是主动行动、做出决策、选择目标、选择客户、选择受众并去服务他们，对吧？但要能利用AI来完成我们周围的工作，而不是害怕它会真正剥夺我们的能动性，因为它做不到，你明白吗？

I think that what I'm most optimistic is I want people to be able to leverage to use their agency more. You know, I think that we have worked so hard for all of our history. And I think we need to stop so focused on like hard, hard, hard work and then being able to like do what we're best at doing, right, which is taking initiative, making decisions, choosing a purpose, choosing a customer, choosing an audience and go serve them, right? But be able to leverage AI to do the work, around us and not be afraid of it that it's really taking away our agency because it just can't, you know?

很棒的公司名。人类拥有能动性。

Great company name. Humans have agency.

确实。太棒了。埃利亚斯，非常感谢你。

Indeed. Awesome. Elias, thank you so much.

谢谢邀请我。

Thank you for having me.

谢谢。

Thank you.