Forward Deployed Engineering at Cursor — Pauline Brunet
频道: AI Engineer
视频: https://www.youtube.com/watch?v=APqXGyCoGW4
原文语言: en
统计: 共 13 轮 · Pauline 13
[0:01] Pauline
[music]
[音乐]
[0:12] Pauline
Hello everyone. Can you hear me okay? Fantastic. Are we having a great conference? Everyone's excited? Amazing. I would love to introduce myself. My name is Pauline Brane. I lead the forward deployment engineering team globally at Cursor. I'm super excited to share some of the learnings that we've had as a company as we build out this function. I have been doing AI deployments to enterprises for the last 10 years across consulting and I've worked in it's my third tech company. So I'm incredibly excited to share some of the learnings. I encourage the discussions. I'm incredibly passionate about the FDE function and only 20 minutes to talk about it is not nearly enough. So please catch me after I'd love to chat with you about how we're thinking about this, what we've learned, and what we want to do going forward. So a couple things about Cursor. We are an AI coding platform as you are well aware. We have incredible products that we are offering as we're helping people go from autonomous coding to asynchronous and synchronous agents all the way to an AI software factory. I'm really going to focus today on the forward deployed engineering function and I want to share some of the learnings that we've had. So I'll really strictly focus on that but always eager to talk to you about what we're doing at Cursor and what we're building. So I always hear about FDE. I'm waiting for the Forbes article that's going to say 2026 hottest job of the year is the FDE. I feel like this is the year and I really want to talk about it. If you're thinking about building an FDE for your business, for your company, what I would personally recommend you do. And the first thing is I want you to understand your customers or at least the target market that you're going after and where they are in their transformation journey. And then I want you to understand your own product. What are you offering to those folks? How customizable is it? And the reason for that is cuz I think of it on on a matrix, right? And a lot of people ask me, "Hey, is FDE like professional services? Is it the same thing as staff augmentation? What is FDE?"
大家好，能听清吗？太好了。这次大会开得开心吗？大家都很兴奋？很棒。我先自我介绍一下：我叫 Pauline Brunet，在 Cursor 负责全球的前置部署工程（FDE）团队。今天特别高兴能和大家分享我们公司在搭建这个职能过程中的一些心得。过去 10 年我一直在做面向企业的 AI 部署，做过咨询，现在是我待的第三家科技公司，所以我非常乐意把这些经验分享出来，也欢迎大家来讨论。我对 FDE 这个职能特别有热情，20 分钟远远讲不完，所以会后请一定来找我聊，我很乐意跟你们聊我们的思路、我们学到了什么、以及接下来想做什么。先简单说说 Cursor：大家都知道，我们是一个 AI 编程平台，有一系列很棒的产品，帮助大家从自主编程走向异步和同步的 agent，最终迈向「AI 软件工厂」。不过今天我会聚焦在 forward deployed engineering 这个职能上，分享我们的经验教训，严格只讲这一块——当然，随时欢迎跟我聊 Cursor 在做什么、在构建什么。我总能听到大家谈 FDE，我一直在等 Forbes 发一篇文章说「2026 年最热门的职业是 FDE」。我觉得今年就是那一年，所以特别想聊聊这个话题。如果你在考虑为自己的业务、自己的公司组建 FDE 团队，我个人的建议是：第一，先搞清楚你的客户，或者至少是你瞄准的目标市场，看他们处在数字化转型旅程的哪个阶段；第二，理解你自己的产品——你到底在向这些人提供什么？它的可定制程度有多高？为什么要这样？因为我是用一个矩阵来思考这件事的。很多人问我：「FDE 是不是就是专业服务（professional services）？是不是跟人力外包（staff augmentation）一回事？FDE 到底是什么？」
[2:16] Pauline
And I have some pretty strong opinions about where it fits and where it's not a great use of your 10X engineers to invest that time in for your customers. So, the first part is think about your digital maturity of your customer. How how mature are they? How technically advanced are they? Where are they on their transformation? And how can you actually help them for it? The second part is the product customization. Like, how configurable is your product? Is it SAS straight out of the box? You just sort of send a login? Like, for example, Teams. Is it super easy to get set up and get running by yourself? Or is it highly customizable, highly configurable for your customers? And think about that on the metric. So, if you have a customer who is really mature in their digital transformation, they have engineers, um it's a low customization on your project, I would really just say provide your product in a self-service fashion, create give great documentation, probably not a great use of your FDE motion. Same thing on your low maturity, low customization. To me, this is like a traditional SAS deployment, right? Like, I go in, deploy it. You can picture the waterfall project going through. Um not a great use of your FDE projects. Then you have customers who are really really mature and you have highly customizable products. Here, you're going to have crazy adoption, right? And I would say a lot of it should be your FDE team should be acting as advisers and help accelerate them. And then finally, you have some of your customers that are sort of further back in their transformation journey. They're learning a little bit about how to do it. They may not be able to hire or staff the folks that you you And so, they need a lot more help and here is where I think of like that embedded transformation. And then across this, that blue sort of square that you see is where I think the FTE really fits in. And for a few reasons, which is for those that are very mature, low customization, you can help them maybe extend, right? Create a couple new features, extend the application. You have that great feedback loop that you give to your product and engineering teams.
关于 FDE 适合放在哪、以及在哪些场景下让你的 10 倍工程师投入时间给客户并不划算，我有一些相当明确的观点。第一个维度是客户的数字化成熟度：他们有多成熟？技术上有多先进？转型走到了哪一步？你能怎么帮他们？第二个维度是产品可定制性：你的产品可配置程度有多高？是开箱即用的 SaaS，发个登录账号就完事——比如 Teams 那种，自己就能轻松装好跑起来？还是高度可定制、可以为客户深度配置的产品？把这两个维度放到矩阵上看。如果客户的数字化转型已经非常成熟，自己有工程师，而你的产品定制化程度又低，那我的建议就是以自助（self-service）的方式提供产品，把文档写好就行，这种场景大概率不值得投入 FDE。低成熟度、低定制化也一样——对我来说那就是传统的 SaaS 部署：进场、部署，你可以想象一个瀑布式项目走完流程，同样不适合 FDE。然后是客户非常成熟、产品又高度可定制的情况：这时候你会看到疯狂的采用率，你的 FDE 团队更多应该扮演顾问的角色，帮他们加速。最后一类客户，他们在转型旅程上走得比较靠后，还在摸索怎么做，可能招不到、配不齐需要的人手，所以需要大量的帮助——这就是我所说的「嵌入式转型」。横跨这几块的、你们看到的那个蓝色方块区域，就是我认为 FDE 真正的位置。原因有几个：对那些很成熟、低定制化的客户，你可以帮他们做延伸，比如加几个新功能、扩展应用，同时你还能把很好的反馈回路带给你的产品和工程团队。
[4:17] Pauline
But your FTE team won't be as helpful or as impactful with those folks. Same thing on high maturity, high customization, there's definitely room to help them, but they can do a lot of their work for them and you don't want to be like a solution architect, you don't want to be writing down bugs, you don't want to be, you know, updating documentation. You really want to focus on what's going to drive business for them. Same thing on the traditional deployment, maybe you can do some configuration extension. I would be very mindful that you're not doing a product 101 201 session, that you're not like holding workshops for everyone in the company to be trained on your latest SaaS software. I think that's not a good use of the FTE function personally. So then, once you've decided, "Hey, where am I in this box?" Hire accordingly, because you have to attract this talent, you have to pay them, and then you have to make sure that they are working on critical, interesting, important things. Otherwise, they're going to get bored, rightfully so, and they're going to leave because you sold them FTE that wasn't FTE in my opinion. So what is a magical unicorn that is the FTE, the forward deployed engineer? To us, it is someone incredibly technical who also has really high IQ. So what does the job entail? It entails working with all types of customers across all levels of the organizations, so CIOs, CTOs, COOs, as well as a developer, engineering managers, working with uh of transformations, VPs of AI. and so you have to be able to really quickly lead a discovery, find the right use case, understand their processes, how are they doing things today, how can you help them do things better. You have to understand their culture, how you're going to affect change, and then you have to accompany them on that digital transformation journey, right? So, you're an actor of change. Of course, you're using technology to do so, but you have to help them accompany them. I always say, if you put in the latest and greatest tech in your organization, and you don't accompany people, no one's going to use it. And then finally, as an FTE, you have to be at the cutting edge of the technological advancements.
但对这类客户，你的 FDE 团队的帮助和影响力不会那么大。高成熟度、高定制化那一格也一样：当然有可以帮忙的空间，但他们自己能干很多活，你不应该只当一个解决方案架构师，不该只是记录 bug、更新文档——你要聚焦在真正能给他们带来业务价值的事情上。传统部署那一格也是，也许你能做一些配置和扩展，但要非常小心，别把自己变成开「产品入门 101、进阶 201」课程的人，别在给全公司办你家最新 SaaS 软件的培训工作坊——我个人认为那不是 FDE 该干的事。所以，一旦你想清楚了「我在这个矩阵里的哪一格」，就照着这个定位去招人。因为你得吸引到这类人才、付得起他们的薪酬，然后还得确保他们做的是关键的、有意思的、重要的事情。否则他们会觉得无聊——这很正常——然后就会离开，因为你卖给他们的「FDE」在我看来根本不是 FDE。那么，FDE（forward deployed engineer）这个「神奇独角兽」到底是什么样的人？对我们来说，这是一个技术极强、同时情商也非常高的人。这份工作具体是干什么的？你要和各种类型的客户打交道，覆盖组织的各个层级——CIO、CTO、COO，也包括一线开发者、工程经理、转型负责人、AI VP。所以你必须能非常快速地主导需求探索（discovery），找到正确的用例，理解他们的流程——他们现在是怎么做事的、你能怎么帮他们做得更好。你要理解他们的文化，想清楚怎么去推动变革，然后陪伴他们走过数字化转型的旅程。所以你是一个变革的推动者。当然你是借助技术来实现，但你必须陪着他们走。我常说，如果你把最新最强的技术放进一个组织，却不去陪伴人、带着人走，没人会去用它。最后，作为 FDE，你必须站在技术进展的最前沿。
[6:26] Pauline
So, I don't know about you guys, but I'm you know, there's like new releases every week. We got to catch this up. You have to be curious. You have to be able to enjoy it and want to learn from it. Otherwise, this is not a fun job, let me tell you, because your customers are expecting you to be the expert, and that's super important. And then finally, you have to work very delicately between the product and engineering teams and your customers, where you want to give that feedback loop to your product team saying, "Hey, we're hearing this over and over again." Uh and the FTE team is so close to the customers, embedded in their organizations, they're going to be the first ones have a really good pulse on what we should build next as a company. So, really effective feedback loop. Uh we work very closely with our engineering and product teams. So, I want to share a little bit about what Cursor looks like uh in terms of the FTE team, what we've learned, what works for us, what doesn't. Uh so, as you know, we are an AI coding platform. A lot of our customers are incredibly [snorts] mature organizations, deeply technical buyers and users, and that has informed of who we hire and the profiles we're looking for. So, I recommend you think about who is buying your software, and how can you help them, and then match the right folks to hire in your FTE team. So, for us, it is project-based, highly impactful projects. So, I usually would like to work with the economic buyer or a very senior champion within my accounts to make sure that we are scoping something that is a strategic objective for the company, right? That is going to drive meaningful ROI for the company, and that they have the resourcing that they're going to put on this project. Because you're going to need to be working with them side by side. You're going to need to get access to their systems cuz we develop on top of their code base. You're going to need to affect change, so you're going to need to have, of course, the top-down support to do so.
我不知道你们怎么样，反正现在每周都有新东西发布，我们得跟得上。你必须有好奇心，必须真心享受并且想从中学习，否则这份工作一点都不好玩——说实话，因为客户期待你就是那个专家，这一点非常重要。最后，你还要在产品工程团队和客户之间非常小心地周旋：你要把反馈回路带给产品团队，告诉他们「嘿，这个需求我们反复听到很多次了」。FDE 团队离客户最近，嵌在客户的组织里，他们会是第一批对「公司接下来该做什么产品」有敏锐感知的人。所以这是一个非常有效的反馈回路，我们和工程、产品团队合作得非常紧密。接下来我想分享一下 Cursor 的 FDE 团队是什么样子，我们学到了什么，哪些做法对我们有效、哪些没有。大家知道，我们是 AI 编程平台，我们的很多客户是非常成熟的组织，买家和用户都有很深的技术背景，这决定了我们招什么样的人、找什么样的画像。所以我建议你想清楚：谁在买你的软件？你能怎么帮到他们？然后据此匹配该招进 FDE 团队的人。对我们来说，FDE 是项目制的、高影响力的项目。我通常希望和客户方的经济决策人（economic buyer）或者一位级别很高的内部支持者（champion）合作，确保我们规划的是这家公司的战略级目标——能带来实实在在 ROI 的事情——并且他们会为这个项目配上资源。因为你要和他们并肩作战，你需要拿到他们系统的访问权限——毕竟我们是在他们的代码库之上做开发——你还要推动变革，所以当然需要自上而下的支持。
[8:28] Pauline
And you're going to make sure that you're working on something really meaningful that when you walk away at the end of the engagements, and we in our case have deployed cloud agents, long-running agents, we've deployed automations, we've built applications on top of our Cursor SDK, that when we walk away, it is a strict ROI for them. That means they're not going to turn things off when we leave, right? Because that's part of the FTE project. We want to make sure that we're affecting change and that we're building things that matter to them. Meaningful return on investment. We also some cool things is we push the edge cases for the Cursor platform. So, as we get really cool use cases that are a little bit outside of the software development life cycle, the FTE team to me is the tip of the spear that is going to test out these new use cases with customers. So, we have incredible folks working with us on, "Hey, how do I help across my HR team? Uh my finance team? My supply chain team? My e-commerce team?" Right? Working with retailers, financial banks. How do I do asset management better? And as we're pushing the the edge use cases for Cursor, we're getting more and more momentum on how we can help you do other things not just within the software development life cycle, which is super important, but also across your entire company. We work in co-development with customer teams in their code base. This is where you have to be really careful that you end up uh not doing staff augmentation. So, if anyone says, uh "Yeah, you have to do this. Uh we're understaffed." Red flag for me personally. I get a little antsy on those phone calls. I'm like, "Ooh, I don't think this is the right case for us." And so, you want to make sure that you're driving something meaningful that they're going to put resources towards and that you're going to work in collaboration. So, my trick is I just ask for who are the people we're going to work with. That's great. I'll be We'd love to partner with you on this use case. We love to build long-running agents to automate your call center ticketing system. Who will be the working team? So, super important to get that.
你还要确保做的事情足够有分量：当项目结束、你离场的时候——我们的项目里部署过云端 agent、长时运行的 agent，做过自动化，也在 Cursor SDK 之上构建过应用——当我们离开时，留下的必须是实打实的 ROI。也就是说，我们走了之后他们不会把东西关掉，对吧？这就是 FDE 项目的一部分：我们要确保真正推动了变革，构建的是对他们真正重要的东西——有意义的投资回报。还有一件很酷的事：我们在为 Cursor 平台探索边缘用例。当我们遇到一些超出软件开发生命周期之外的很酷的用例时，FDE 团队在我看来就是「矛尖」，负责和客户一起去验证这些新用例。我们有非常出色的伙伴在跟我们一起探索：「怎么帮我的 HR 团队？财务团队？供应链团队？电商团队？」对吧？我们和零售商、金融机构合作——怎么把资产管理做得更好？随着我们不断推进 Cursor 的边缘用例，我们越来越有势能去帮助大家做软件开发生命周期之外的事情——这当然很重要——而且是覆盖整个公司的。我们和客户团队在他们的代码库里做共同开发（co-development）。这里你必须非常小心，别一不留神变成了人力外包。如果有人说「你们得来干这个，我们人手不够」——对我个人来说这就是红灯警报。接到这种电话我会有点坐立不安，心想「哦，这恐怕不适合我们」。所以你要确保推动的是有分量的事情，客户愿意为之投入资源，而且是真正的协作。我的小技巧就是直接问对方：我们会和哪些人一起工作？「太好了，我们很乐意和你们在这个用例上合作，我们很想构建长时运行的 agent 来自动化你们呼叫中心的工单系统——那么，具体的工作团队会是谁？」问清楚这一点非常重要。
[10:21] Pauline
And then finally, working between our engineering and our product teams to influence the road map um and our customers. So, let them know new things are coming that are going to all of a sudden enable use cases we couldn't do before. And that's the beauty of it is we can actually go and solve more things at scale very quickly. I would recommend you have a mission of the FTE. I'll offer you mine. This is the Cursor FTE mission. Uh we partner with your organization to co-design and co-build your AI software factory. We transform how you design, develop, and maintain software across your entire life cycle. I would just encourage that you have one. It's very clear to the customer what you're trying to drive. It's clear to your team of what we should focus on. And it helps our team members think, "Hey, when I'm hearing this project that sounds like staff augmentation, I don't feel like I'm doing that towards our mission." So, that's ours. I offer you your own. What does the team structure at uh Cursor look like for the FTE is we are highly technical, highly experienced profiles. For the beginning, it makes sense for you to hire what I call unicorns. So, we hire 5 plus years software engineers. We don't hire out of school. We don't hire early career professionals at this moment. Once we grow the team bigger, we will. And so I welcome those folks that are reaching out to us. For the first set of founding for deployed engineers, we are hiring very technical folks with customer-facing experience. Over time, we will split the role, so we'll have folks that are a little bit less technical, more customer-facing, but still with technical aptitude and vice versa. Those are highly technical folks who perhaps are not as customer-ready, but have the aptitude to learn it. We have a matrix organization. We are ready to pivot in any way, shape, or form. That will happen. We are geography-based for now, but at some point we will likely mature to industries because when you talk to an industry and you don't use their lingo, you immediately lose credibility. So if you talk to a bank and you're not talking about payment systems, if you're not talking about asset management, risk, you kind of lost them.
最后一点，是在我们的工程团队、产品团队和客户之间穿针引线，去影响产品路线图，同时让客户知道：新东西要来了，它会一下子解锁我们以前做不到的用例。这也是这件事的美妙之处——我们能非常快速地、规模化地解决更多问题。我建议你给 FDE 团队定一个使命（mission）。我把我们的分享给你，这是 Cursor FDE 的使命：我们与你的组织合作，共同设计、共同构建你的 AI 软件工厂；我们变革你在整个生命周期中设计、开发和维护软件的方式。我只是想鼓励你也定一个：它让客户清楚你要推动什么，让你的团队清楚该聚焦什么，也帮团队成员做判断——「嘿，我听到的这个项目怎么像是人力外包？这不符合我们的使命。」这是我们的版本，你可以定你自己的。那 Cursor 的 FDE 团队结构是什么样的？我们招的是技术很强、经验很深的画像。在初期，招我称之为「独角兽」的人是合理的：我们招 5 年以上经验的软件工程师，现阶段不招应届生、不招职业早期的人——等团队规模变大之后会招，所以也欢迎那些正在联系我们的朋友。第一批创始 FDE，我们招的是技术非常强、同时有面向客户经验的人。随着时间推移，我们会把这个角色拆分：一部分人技术稍弱一些、更偏客户沟通，但仍有技术素养；反过来也一样——技术非常强、可能还没那么擅长面对客户，但有学习的潜质。我们是矩阵式组织，随时准备以任何方式调整——这一定会发生。目前我们按地理区域划分，但未来很可能会走向按行业划分，因为当你和一个行业对话却不说他们的行话时，你会立刻失去可信度。比如你跟一家银行聊，却不谈支付系统、不谈资产管理和风控，你基本上就把他们聊丢了。
[12:25] Pauline
And so you really have to have that industry knowledge. And then finally, across our product areas, we really want to focus on having the right SMEs who then become the experts. So for example, we have someone on the team who is the expert on our long-running cloud agents. We have someone who's an expert on the cursor SDK. And other team members can come in on their projects and say, "Hey, I need to pull in this person because they're important." So for us, we have folks phenomenal folks from Spotify, from Rippling, from Palantir, from a lot of organizations where we have found that we get the best people because they are incredibly excited to work with customers and they have the aptitude to do so. And then finally, we will change the roles and the structure over time. One thing we always joke about on my team is what we're doing today is not what we're going to do 6 months from now. We won't be hiring for the same profiles. We'll have changed. New products will have come out. Our customers will have changed in their journey. Here are some best practices. By the way, we We in no way perfect. I have done this for 10 years. I've made a ton of mistakes. Um learn quickly and pivot is my recommendation. Just learn from it. Try out a project with a customer. You might fail. That's okay. Right? Just actually try it out. You will learn so much more than if you're just kind of waiting and planning. I don't recommend doing that. Listen to your customers. I'll give you a really concrete example. Um I was not planning on offering this as an FTE offering. I've heard it six or seven times now and so I'm going to create one. The question I get asked is, "Hey Pauline, how do I change my organization now that we have these amazing tools? How do I capture value? So, who do I hire? What are the job description? Um how do I rearrange the teams? Right? How do I change the ways of working together, the processes to actually go capture this value?" Not something I was going to offer. Um so, we actually might be offering that very soon and we'll hire the right team to actually support that. So, listen to your customers and adopt accordingly. Uh work with partner organizations.
所以你必须真正具备行业知识。最后，在各个产品领域，我们希望配置对的领域专家（SME），让他们成长为专家。比如团队里有人是我们长时运行云端 agent 的专家，有人是 Cursor SDK 的专家，其他成员在自己的项目里可以说「嘿，我需要把这个人拉进来，因为他很关键」。我们团队里有来自 Spotify、Rippling、Palantir 等很多组织的出色人才——我们发现最好的人，是那些真心乐于和客户打交道、并且有这个能力的人。最后一点：角色和结构会随时间不断调整。我们团队常开玩笑说，今天做的事和 6 个月后做的事不会一样：招的画像会变，新产品会发布，客户也会走到旅程的新阶段。接下来是一些最佳实践。顺便说一句，我们绝不完美——这行我干了 10 年，犯过一大堆错误。我的建议是：快速学习、及时调整，从错误里吸取教训就好。和客户试着跑一个项目，可能会失败，没关系，对吧？真正动手去试，你学到的会远远多于坐在那里等待和规划——我不建议干等。倾听你的客户。给你们一个非常具体的例子：有一项服务我原本没打算做成 FDE 的产品，但现在已经听到六七次同样的需求了，所以我准备把它做出来。大家问我的问题是：「Pauline，现在有了这些强大的工具，我该怎么改造我的组织？怎么真正把价值收进来？该招什么人？职位描述怎么写？团队怎么重组？协作方式、流程要怎么改，才能真正把这些价值拿到手？」这本来不在我的计划里，但我们可能很快就会推出这项服务，并招到合适的团队来支撑它。所以：倾听客户，随之调整。还有，和合作伙伴组织一起做。
[14:27] Pauline
You can always benefit from uh system integrators and consultings. One, they really know your customers. They've been in there for a while. They have great relationships. Two, there's a lot of stuff that I just frankly don't want to do and let's have the SIs do it. So, for example, I'm not really that great at change management. Not a fan favorite. Um and so, with a lot of that can be actually accompanied with partners. Same thing on rolling out uh existing products that we're really good at and we've done this a bunch across Telco. We've done it a bunch across healthcare life sciences. You can roll this out and increase the reach of your organizations. So, partner with your your uh system integrators and your consultants so they can go do that at scale and broaden in your reach. Talked about this already. Um don't be afraid to say no. This one's a hot topic. Uh I have customers who'll say, "I want to do this." And I will say, "Ooh, Cursor is not the right tool for that." Couple of reasons. One is you build credibility by being very honest about where our applications and our products and our platform are the right tools and where they're not. So, you earn a lot of um sort of feedback from that because they'll say, "Actually, I have this other use case now that I trust you as a trusted partner." And then two, of course, if if it doesn't go well, then you're on the hook for that. So, I would recommend be very specific in the use cases you're solving. And then finally, attract and pay the right talent. So, just make sure that you are hiring the right profile for what you want to deliver on. Very dependent on you. We talked about this at the beginning. And make sure that you are attracting them, paying them, keeping them motivated. So, very important, especially in this you know, war for talent that we're currently in. Here are the tips for running through FTE. Uh check you're solving the right problem. It seems easy. It's actually really not. Uh sometimes you're you're solving sort of a symptom, not the right problem. Sometimes you're talking to the wrong person who thinks they have the right problem, but they may not. So, just always inquire, ask questions, um who's responsible for this?
系统集成商（SI）和咨询公司永远能帮到你。第一，他们真的懂你的客户——在客户那里深耕多年，关系非常好。第二，有很多活儿说实话我自己不想干，那就交给 SI 去做。比如变革管理（change management），我不太擅长，也不是我喜欢的活儿，这类工作很多都可以交给合作伙伴一起做。同样，推广我们已经很成熟的产品也是——我们在电信行业做过很多次，在医疗和生命科学领域也做过很多次，完全可以让他们去铺开，扩大你组织的触达范围。所以，跟你的系统集成商和咨询伙伴合作，让他们规模化地去做这些事，帮你放大覆盖面。这点前面提过了：不要害怕说不。这是个热门话题。有客户会说'我想做这个'，我会直接说：'哦，Cursor 不是干这个的合适工具。'原因有两个。一是，坦诚地讲清楚我们的应用、产品和平台在哪些场景合适、哪些不合适，能建立信誉。这样反而会换来很多回馈——客户会说'既然你是值得信赖的伙伴，那我其实还有另一个用例想找你'。二是，当然，如果项目做砸了，锅是你来背的。所以我建议：对你要解决的用例保持非常明确的边界。最后一点，吸引并给对的人才付对的钱。确保你招的人画像和你要交付的东西匹配——这非常取决于你自己的情况，开头我们聊过这点。还要确保你能吸引他们、给足薪酬、让他们保持干劲。这非常重要，尤其是在当下这场人才争夺战里。接下来是做 FDE 项目的几条实操建议。第一，确认你在解决对的问题。听起来简单，其实真不容易。有时候你解决的只是症状，不是真正的问题；有时候你聊的对象自以为抓住了问题，但其实没有。所以要不断追问：这事到底归谁负责？
[16:25] Pauline
Can I talk to them? Right? I really like to talk to the person responsible for the process, the workflow, the department, whatever we're trying to solve for. Define success from the start. If I accomplish this, will this be successful for you? If I automate this process from start to finish and then now takes 20 minutes instead of 3 hours, which is current baseline, is that sufficient for you? Does that measure success? Yes, great. Um I I have a lot of opinions about scope. So, I think that you cannot just do like, "Hey, take two FTEs for 6 months, do whatever you want with them." I think that's a recipe for failure. What I would recommend is that you instead are trying to solve a problem and you establish what you're going to do to solve that problem. So, we want to automate this process from our start to finish using long-running agents. Long running agents are going to grab data from these. They're going to make these decisions. They're going to involve these people in the feedback loop, and they're going to go and reduce our mean time to resolution. They're going to go and automate this process from start to finish. They're going to reduce in terms of claims management the time to answer a customer on their claim. Whatever the KPIs that we're trying to drive, keep the scope directional. We're going to do phase one and two. We're going to automate these things. We're going to set up these agents for you. It's going to take 6 weeks. We're going to do as much or as little as we can. The reason for that is I do not know the customer processes. I haven't really seen their data. I haven't seen their systems, right? And so I'm a little bit on the hook if it takes more than 6 weeks or less. That's the first part. The second part is from a customer perspective, we're going to learn a lot, and they're going to maybe want to pivot once we learn something. And having something that's directional where I can go and sort of pivot based on the the learnings I have, customers actually really appreciate that in the end. And so that would be my recommendation. Involve the customer in every step step.
我能跟负责人聊聊吗？我特别喜欢直接找那个对流程、对工作流、对部门真正负责的人谈，不管我们要解决的是什么。第二，从一开始就定义什么叫成功。'如果我做成了这件事，对你来说算成功吗？如果我把这个流程从头到尾自动化，从现在基线的 3 小时缩短到 20 分钟，这够不够？这算达标吗？'——'算，很好。'关于 scope（范围界定），我有很多自己的看法。我认为绝对不能搞成'嘿，给你两个 FDE，用 6 个月，随便你怎么使'——那是失败的配方。我的建议是：你要围绕一个具体问题，把解决方案定下来。比如：我们要用长时运行的 agent 把这个流程端到端自动化。这些 agent 会从这些系统抓数据、做这些决策、把这些人拉进反馈闭环，然后去降低我们的平均修复时长（mean time to resolution），把流程从头到尾自动化；在理赔管理场景里，就是缩短回复客户理赔的时间。不管要驱动的 KPI 是什么，让 scope 保持'方向性'：我们做第一、第二阶段，自动化这些环节，帮你搭好这些 agent，用时 6 周，能做多少做多少。为什么这么定？第一，我并不了解客户的内部流程，没真正看过他们的数据和系统，如果承诺死了 6 周，超了或者提前了我都要担责。第二，从客户角度看，过程中我们会学到很多东西，学到之后他们可能想调整方向。有一个方向性的框架，让我可以根据学到的东西灵活转向——到最后客户其实非常欣赏这一点。所以这是我的建议。还有：让客户参与每一步。
[18:27] Pauline
Scoping, actually building and designing the solution, implementation, doing the human in the loop validation, looking at baseline versus the results, identifying the ROI we're driving. They should own that. We are supporting them in this journey, right? We are still hands-on keyboard. We are still configuring. We are still developing on top of their code base, but make sure you're not doing it alone. If you're doing it alone in their office in a little cubicle, we have a problem. Okay? Please raise your hand if that's the case. We need to solve something. Finally, measure success. Did were we successful? I said that we could do this from 3 hours to 20 minutes. Did we get close? If not, why not? What can we do about it? And then finally, what's the return on investment? You want to over communicate that. Right? I had a someone mentioning to me that an agent was running was costing $2,000 per day and I said, "Well, what was the agent doing?" And he explained it to me and I said, "Hey, that sounds like you're actually reducing costs to send the right person to go fix this equipment. Would that not be worth $2,000 a day?" And he said, "Absolutely, but I never measured it this way." So, always think about what's the ROI you're driving. It's always three things, super simple. Am I increasing revenue? Am I decreasing costs? Or am I mitigating risks? That's it. Every company, as complex as they are, that's what they care about. Which one are you doing? Could be all three, which is fantastic, but at least one of them. Finally, leave your documentation artifacts behind so that you can help them. So, I'll 32-seconds wrap up. Build the right FTE motion for your company. Learn, pivot, and scale, super important. And then create this amazing culture where people want to be a part of it. They want to learn. They want to do their best work. And give them the chance to do so. And so, the the thing I'll leave you with is um you know, I always say hire A players, hire A players, B players hire C players. So, just be very mindful of that. Make sure that you are hiring the right talent for your organization, for your customers, and you're keeping them motivated.
范围界定、方案的设计与搭建、落地实施、human-in-the-loop 的人工校验、基线与结果的对比、算清楚我们创造的 ROI——这些都应该由客户来'拥有'，我们是在这段旅程里支持他们。我们仍然亲自上手写代码、做配置、在他们的代码库上开发，但千万别一个人闷头干。如果你是一个人窝在客户办公室的小隔间里单干，那就出问题了——真是这样的话请举手，我们得聊聊。最后，衡量成功。我们成功了吗？我说过能从 3 小时干到 20 分钟，接近了吗？没有的话，为什么？还能做什么？然后，投资回报（ROI）是多少？这一点要反复地、超量地去沟通。有人跟我说，一个 agent 跑起来每天要花 2000 美元。我问：'这个 agent 在干什么？'他解释完，我说：'听起来它是在帮你把对的人派去修对的设备、降低成本——这难道不值每天 2000 美元吗？'他说：'绝对值，但我从来没这么算过账。'所以，永远要想清楚你在驱动什么 ROI。说到底就三件事，超级简单：我是在增加收入？降低成本？还是在缓释风险？就这些。再复杂的公司，关心的也就是这三样。你在做哪一样？三样全占当然最好，但至少得占一样。最后，把文档和交付物留下来，方便客户后续使用。好，用 32 秒收个尾：为你的公司搭建适合自己的 FDE 模式；学习、调整、再规模化，非常重要；然后营造一种让人向往的文化——大家想加入、想学习、想做出最好的工作，你要给他们这个机会。最后送大家一句我常说的话：A 级人才会招来 A 级人才，B 级人才会招来 C 级人才。所以一定要注意：为你的组织、为你的客户招对的人，并让他们保持干劲。
[20:29] Pauline
Thank you, everyone.
谢谢大家。
[20:45] Pauline
[music]
[音乐]