AI赋能产品团队的契机
在一次深刻的体验之后，我立即意识到需要提升整个产品团队的技能水平，不仅是设计师，还包括产品经理。我们需要更熟悉 AI技术（Artificial Intelligence: 模拟人类智能的机器系统），并理解如何运用它。我休陪产假时发出的这条信息，成功激发了领导们的极大热情。我当时并不完全清楚事情会如何发展，只知道需要让更多人关注 AI。主持人指出，如果你是第一个站出来表示要探索团队如何利用 AI、并愿意领导这一变革的人，这将是一个独特的领导机会，能够展示跨职能的广泛影响力。
Original English
I had tried Cursor for the first time, and what I was able to create just blew me away. I sent a message to my manager, my manager's manager, the CPO, and then a few other folks that I knew were really interested in AI, and I was like, "Listen, I had this really profound experience, and I think we really need to uplevel the skill of our entire product organization, not just designers, but also PMs. We need to become more familiar with this technology. We need to understand how we can use it." This is actually the message that I sent while I was on paternity leave that definitely got my leaders really fired up. I didn't know exactly how this was going to go. All I knew was that I needed to get more folks paying attention to this AI stuff. If you were the first to raise your hand that says, you know what, I want to figure out how our team can use AI. I'm going to lead this organization. It's such a unique leadership opportunity to show cross-functional broad impact on teams.个人经历：AI如何改变工作流
我是 Pendo 的一名设计师，一直密切关注 AI 领域。去年11月，我首次尝试了 Cursor（AI-powered code editor: 一款由AI驱动的代码编辑器），它所能实现的功能令我惊叹不已。我有一个关于音乐播放器的业余项目想法：通过扫描二维码播放专辑，以弥补没有唱片机的遗憾。我不是一名活跃的开发者，无法独立编写这样的移动应用。然而，使用 Cursor 仅几个小时，我就创建了一个可工作的原型，能够生成二维码和PDF。我立即意识到，除了娱乐项目，AI 还能用于构建交互式原型。虽然我精通 Figma（Design tool: 一款流行的UI/UX设计工具）原型设计，但它在处理数据驱动的分析功能时存在局限。而基于代码的动态原型，即使使用模拟数据，也能更好地展示真实数据下的功能效果。因此，我深信 Pendo 也能从中受益。
Original English
Yeah, absolutely. So, to take you back, it's the end of last year. So, last July, I had my daughter Maya was born. And I was on paternity leave at the end of last year, and I'm sort of like a tech geek. I've been following AI for a while. Like my day job is a designer at Pendo, but you know, I've always sort of been into tech, and so been following AI very closely. I've also had some experience like building side projects and things like that. And I think it was back in November, Cursor came out, or maybe it was a little bit earlier than that, but Cursor I had tried Cursor for the same for the first time. And what I was able to create just blew me away. So, I had like a side project idea, this hobby app in my in my mind about a music player where I can play albums by scanning a QR code on on sort of like a a piece of paper. I was very jealous of people who had record players. I don't have space for a record player in New York, and they get to choose music by sort of flipping through through albums, and I like having you know unlimited access to to music via Spotify, but I sort of miss that sort of like tactile experience. So, I was like what if I create these sort of like laminated cards and I have this way of just being able to play the albums that are on that. So, I had this idea of like what if I can create sort of a mobile app where I can scan like a QR code or I could recognize the album cover and I can print these album covers out and just be limited in in that way. And I I had no idea how to do that like on my own. Like I'm not an active developer. I can't sit down and write that application. And I pulled up Cursor and like within a couple hours I had a working prototype. And like that just blew me away. I was creating QR codes. I was creating PDFs. I was like doing all this like really really really cool stuff. And you know like I said my my day job is is a product designer. And I immediately understood that like, okay, this is really cool as a sort of side project. That's really fun, but I could use this to build interactive prototypes. I'm I'm I consider myself pretty proficient with Figma, especially when it comes to prototyping, but I understand the limitations of using Figma for prototyping. A lot of what I do at Pendo is sort of working on features that are analytics based. And so, when you're creating mock-ups and prototypes that are data driven, it's really hard to communicate what the, you know, how these things are actually going to work with real data. So, having a prototype that is code-based that is working with even just fake data and interacting sort of in a more dynamic way is really useful. So, I was like wow I we could really use this at Pendo.推动AI转型的双重目标
尽管我当时还在休陪产假，但我无法抑制住内心的想法。于是在12月，我通过 Slack（Team communication platform: 一款团队协作和沟通平台）向 Pendo 的多位同事发送了一条长消息，包括我的经理、经理的经理、首席产品官（CPO）以及其他对 AI 感兴趣的同事。我分享了我的深刻体验，并强调我们需要提升整个产品团队的技能，让设计师和产品经理都熟悉这项技术及其应用方式。我意识到，学习 AI 没有现成的“剧本”或课程，技术发展如此迅速，唯一有效的方法就是通过实践来熟悉其工作原理，紧跟最新技术和工具，并从大量案例中学习。
我提出的 AI 转型倡议有两大目标：首先，让 Pendo 的产品团队能够利用尖端 AI工具（AI Tools: 辅助完成特定任务的AI应用）以更少的时间和资源完成更多工作，改进决策，并更有效地沟通和验证想法。其次，由于 Pendo 服务于许多正在经历类似转型的产品组织，此举也能帮助 Pendo 在该领域树立 思想领袖（Thought Leader: 在特定领域拥有深刻见解和影响力的专家）的地位。我的领导们对此非常兴奋，CPO甚至邀请我在全员会议上分享这个概念。虽然我当时还在休假，但承诺复工后立即启动。
Original English
So I had the, you know, even though I am on paternity leave, you know, I had this idea and I was like I couldn't contain myself, you know, I had I wasn't going to come back to work until the beginning of January. So, in December, I wrote a whole bunch of folks at Pendo. I still had access to my Slack. And so I wrote, you know, I sent a message to my manager, my manager's manager, the CPO, and then a few other folks that I knew were really, really interested in AI. And I was like, listen, I had this like really profound experience. And I think, you know, we really need to uplevel the skill of our entire product organization, not just designers, but also PMs. We need to become more familiar with this technology. We need to understand how we can use it. I had already understood that like there's no playbook for how to learn this stuff. There's no class you can take. There's no book you can read. And the technology is evolving so fast that the only way to really know how to apply it is to become very familiar with how it works to kind of stay current with all the latest technologies and the tools and just sort of like see a bunch of examples and like selfishly like I wanted to spend more of my time at work doing these things, but I also wanted to help my my colleagues and my company just be more successful because I I saw a clear path to that and just getting more pe more my peers, more of my colleagues like doing the same thing and sharing their experiences. is I know would help me learn and I think it would just sort of like you know rise all the boats and so this is sort of an example here not example this is actually the message I pulled it up that I sent while I was on paternity leave just to kind of give you an example of what something like this looks like and I wrote this like pretty long message I mean for folks that are not watching this you know I just kind of said TLDDR in a similar fashion to you know there's an and there's an engineering focused sort of group that had that had been around for at least a here but nothing really focused on PMs and designers. I was like I'd like to lead a group like that before a cross functional cross functional product team with design designers PMs and etc with two goals. Pendo's product team can leverage the cutting edge of AI tools to get more done in fewer hours and less resources improve decision-m and communicate and validate ideas more effectively. And then two, because Pendo is also sort of servicing a lot of product organizations that are going through similar transformation to help position Pendo as a thought leader in the space because I knew it was going to be really important. And then I went through like a longer version of sort of like my my experience of like building this app and and why I thought it was important. And you know I'm really fortunate to be part of an organization that supports sort of initiatives like this. And so I, you know, that that definitely got my my leaders like really fired up. In fact, the CPO was like, "Hey, can you come to like all hands next Monday and talk about this this concept?" And I was like, "I'm on paternity leave. I can't do it yet." But I will start as soon as I come back. And so that that's sort of that's sort of what happened. So that was a catalyst for this idea. And I gotta say like I'm I'm the kind of person that you know sometimes I can be like type A and like really like think things through, but I also know that sort of committing to something or just like forcing myself to like throw myself into a situation without knowing like how it's going to work out can also result in something really interesting. So I didn't know exactly how this was going to go. All I knew was that I needed to get more folks paying attention to this AI stuff, and I also needed to create time in people's calendar where everyone could just like focus on it and play or maybe hear a presentation on on something new.AI转型：职业发展的独特机遇
主持人强调，对于任何希望在组织内推动 AI转型（AI Transformation: 组织通过整合AI技术实现业务流程和文化变革）的人来说，这是一个重要的机会。Brian提出的两个核心点——团队需要掌握 AI工具 以提高效率，以及组织可以借此成为行业内的 思想领袖——与 LaunchDarkly（Feature management platform: 一家提供功能管理平台的公司）的 AI转型 经验不谋而合。这不仅是技术进步的必然，更是公司在未来软件工程领域保持领先的关键。
此外，主持人指出，这种 AI转型 倡议也是一个独特的职业发展机遇。作为第一个站出来领导团队探索 AI 应用的设计师或产品经理，你将有机会展示跨职能的广泛影响力，并成为这一领域的领导者。这不仅对团队有益，对个人职业生涯也大有裨益。Brian也认同，这项倡议确实为他打开了许多内部机会，让他能够参与酷炫的 AI项目，并被同事视为 思想领袖，影响力远超其日常工作范围。
Original English
So, I have to call out a couple things here that I think are really important. One for anybody trying to, you know, give a a a justification if you if you need it for investing extra time, resources, and energy into this AI transformation in your organization. I love that you call out actually the two things that really matter. They're very similar things to how I called out the value of AI transformation at LaunchDarkly, which was one, our team's got to know how to use this stuff. Like we've just got to know how to use these tools, get more done, be more efficient, just use the best of the best. The second one though I think is really interesting and there's still a lot of opportunity here. You know, you're on this podcast which is there's this opportunity for leading organizations to position themselves as thought leaders in how you get stuff done with AI in your vertical. And so for us it was like we have to be great AI engineers because we need to you know be great engineers generally. This is the next phase of how software engineering is going to get done. We need to be thought leaders in the space. And very similarly for you on the product side. I think it's just really important that you can create platforms for your company to be experts in the space if you lean in early into these technologies. You know, the other thing I want to call out is, you know, I try to tell people about this all the time. This is like promo making work. And what I mean is like this is the kind of initiative that doesn't come around that often as an opportunity. And if you are the first to raise your hand, like if you are the first designer that says, you know what, I want to figure out how our team can use AI. I'm going to lead this organization. It's such a unique leadership opportunity to show cross-functional broad impact on teams and like there's only gonna be one or two of you that get to be the leader of it. So I'm like really encouraging people to be like you raise your hand early to take on the initiative for the organization. One because I think it's the right thing to do for the team, but two, it's really great from a personal career perspective. Absolutely. And like you know I wasn't I wasn't going to focus on that. But I will but like what you're saying is absolutely 100% true. So like this this sort of initiative has opened so many doors. We'll get into it in a moment but it's opened so many doors internally within the organization. Like I'm speaking with you. There's no way I think I would be speaking on such a high-profile podcast if I didn't start working on this and sort of build up a sort of the the body of work that I have over the last nine months. I get to work on some really cool AI projects. I have folks throughout the organization that are not even in product and and and and design, folks I didn't even know reaching out to me and sort of like looking to me as a thought leader. And that wasn't my intent, but it's absolutely true. It's like there there are opportunities across all organizations right now regardless of your level. I mean, I'm a senior staff but I'm an IC. I'm I'm just a product designer like but I'm having an influence way beyond my my scope. And I think regardless of where you are like if you have the initiative and the energy and it does take a little bit of time like there is some nights and weekends that like I kind of put into it but I also love this stuff like I was kind of doing it anyway and so absolutely it is it is a career builder builder.推动AI采纳的双管齐下方法
回到工作岗位后，我立即在 Pendo 内部的产品组织（包括产品经理、设计师、文案等）中宣布了这项 AI 倡议。我采取了“双管齐下”的方法：异步沟通（Asynchronous Communication: 非实时进行的沟通方式，如邮件、Slack消息）和 同步会议（Synchronous Meeting: 实时进行的会议，如视频会议、面对面会议）。许多人会说“我知道 AI 很重要，但我没时间学习”，但如果不投入时间，就永远无法掌握。因此，除了在 Slack 上鼓励大家异步分享和学习，更重要的是在日历上为团队成员预留出专门的时间，让他们能够专注于 AI 学习和实践。
这些同步会议不仅是演示，更强调互动性。例如，在一次启动会议中，我们进行了一个构建应用的练习。我强调 AI 如何加速软件的规划、设计和构建。我引用了 Andrew Ing（AI researcher and entrepreneur: 一位著名的AI研究员和企业家）关于产品经理在 AI 时代定位的观点，他认为产品经理和设计师都需要具备技术熟练度、迭代开发能力、数据技能、管理模糊性和持续学习的能力。
Original English
Yep. Yeah. So, the two things, so I'm just kind of bringing up here in case it's helpful. So, this was sort of the first announcement I made within the organization had come back in January. We this is sort of like our private channel of the entire product organization. So, within Pendo product the product team are PMs, designers, writers, and a few other folks. And so I was like, hey, I'm starting this initiative. And in my mind, I I I was thinking that there's sort of a two-prong approach. There's an asynchronous and a synchronous. One thing that I was very familiar with, like just in my own life, but also talking to other people, is that you'll typically hear something like, "Yeah, that AI stuff like I know it's important, but I just don't have the time." Like I don't have the time to like watch all the videos and you know, vibe code and lovable or whatever it is, right? And the crazy thing is like if you don't make the time for it, you're never going to learn it and at some point you're going to get behind, right? And so it was really important for not just there to be a place within Slack and encourage people to share on Slack asynchronously to do it at their own pace, but also to create time in people's calendars so that they can come and focus on like whatever the topic is. And also within that session, it's not just about a presentation. It's also like it's really important for that presentation or that session to be interactive. So, let me give you an example of that. So, this was this was a this was a kickoff as well as an exercise about building apps. Just to sort of give you a insight into how I start started this off. I was like hey AI is getting better. It's getting faster and it's evolving how software is planned, designed and built. And this was really meant to speak to not just us as builders but also sort of the the product that Pendo was building. And right around that time, Andrew Ing had a a really I thought thoughtful blog post about how he was thinking about how PMs need to position themselves sort of in the AI in in this AI future. And then AI is typically sort of in the space of sort of engineering and technical stuff and you know he was positioning AI as perhaps being or at least engineers being better positioned to sort of like take advantage of all this stuff because they are technical. But he's saying it was really important for PMs and I would also put designers in that camp as well to become proficient. And so these were the sort of five things that he was he really wanted to focus on technical proficiency just iterating on the development like using AI in a sort of iterative capacity being very proficient with data skill and managing ambiguity and then ongoing learning. So like that was really sort of the emphasis I wanted to drive.互动实践：激发创造力与突破MVP思维
我们 产品AI 倡议的目标是提升和更新产品人员的技能，提高他们的理解力和素养，并帮助客户解决问题。在启动会议中，我强调了互动实践的重要性。我让所有人在 Bolt.new（AI app building tool: 一个利用AI构建应用的工具）上创建一个待办事项列表应用。大家输入相同的 提示词（Prompt: 引导AI生成内容的指令），然后点击“增强提示”按钮，结果却各不相同，甚至有些出现了错误。这展示了 AI生成式（Generative AI: 能够生成文本、图像等新内容的AI模型）应用的多样性以及处理错误的重要性。
在接下来的10-15分钟里，我鼓励大家进行“疯狂”的实验，比如让应用添加“复古8位像素艺术主题”或“MySpace 2007风格”。这种做法旨在让设计师和产品经理体验到 AI 的乐趣，拓宽他们对技术应用的想象力。主持人也指出，这种实践有助于设计师和产品经理摆脱长期以来被“范围蔓延”所束缚的 最小可行产品（Minimum Viable Product, MVP: 具有最少功能但能满足用户核心需求的产品）思维，重新点燃构建“卓越产品”的激情，而不是仅仅满足于“可行产品”。 AI 让实现这些“神奇”功能变得更简单、更高效。
Original English
And so again the structure of sort of like or at least the goal sorry the goals of the the product AI was around to uplevel and modernize the skill set of our of our product people to improve our comprehension and literacy and then because our customers are also builders to empathize and assist with them as well. And I wasn't sure if this technique again was going to work but this was how I was thinking about approaching it. Just being more hands-on, getting our hands dirty, radical many to many sharing, being intentional, creating the space, and then identifying the patterns that worked, and then sort of turn those into reusable patterns. So, in this opening session, like I was saying, like it's really important to have an interactive session. And I'm not going to go through the slides of like how I talked about code building tools and the different, you know, the various types and some things that I had built. But there was a section that was at least I think 10 15 minutes where I was like, "All right, everyone, you know, go to bolt.new." That was the app that we chose to use at the time and create an account if you haven't created an account. And then everyone go to Bolt and copy and paste this. So, this was just a prompt that I had created. It wasn't hyper optimized. It was about creating a to-do list, the most basic like little mini SAS app that you could possibly build. And I had everyone type this in. And then there was like a little enhanced prompt thing that I wanted everyone to use just to sort of see how like this this app could sort of take a very basic thing and and turn it into a more sophisticated thing and then let it rip and hit go. And I would say I wasn't expecting this when I did this the first time, but the thing that really stood out because it was sort of like obvious to me that this is this is what would happen. But the some of the feedback I got was like wow like we all typed in the same thing. We all clicked on the enhance prompt button and we all got different results. So like this was just sort of an example of like these were all the to-do list applications that the app created after running that that query and I think in in a third of the cases like Bolt just came back and said nope like like like error whatever and then like so we immediately got into sort of the whole oh okay don't worry like if you get an error just to fix it blah blah blah and everyone sort of like got through it after two or three rounds. So that is like it was great to experience that as a team and not only just to like do it and see how it worked but also to see the diversity of like how like these applications are built and how Genai is being used and then the other thing I had people do in the last like 105 minutes is experiment on their own and I told people just to do crazy stuff like we're not doing this for any like we're not really building a to-do list and the AI will do its best to do as you say so you can give it like the most wacky you know instructions right like, you know, make it, you know, add a retro 8bit pixel art theme, you know, introduce a dark mode title, make it look like MySpace from 2007. And so that was sort of like an interesting thing, too. So, like people sort of like went nuts. They shared some things like I think there's a few examples here of like people saying, "Oh, you know, I tried. Yeah, here here this is what UI will look like in 20 uh 200. Here's my Tumblr style, you know, to-do list." And like this was intentional to sort of like make it fun. These are designers, but there's also PMs and like we all know sort of like the there's a line between sort of like professional stuff and personal stuff. But like what I really wanted people to experience is that this can be fun and just like to broaden their minds about how they can apply this technology. One thing I want to call out as a meta benefit to this slide that you showed and maybe we could go back to it of like now go wild or or optimize it is I think as designers and product managers in companies and you can tell me if you have a different experience but we have just gotten beaten by the scope creep stick so frequently that we have actually lost our muscle for like asking for the magic thing. We always start with the MVP. We always start with like what is the bare minimum thing I can ship to meet the user requirements that I know engineering can do and like we've lost this ability to imagine like what if it did this and what if it did that and it could be interactive or there could be voice and what I like about AI is one it makes those magic things a lot easier to build you know more efficient to build but two like it's going to let designers and product managers return to the craft of building the awesome product as opposed to like the viable product which is so like if you reflect on it it's so sad that we have put on a pedestal like minimum viability as just like such a low bar and now our bar can just be so much higher for what we build but you have to like reignite this muscle of like how to even think about what those things could be. So, I love the idea that you, you know, put these these iterations in categories like visual iterations, interactive iterations, entertaining or gamification iterations, and then like media again is something that's really interesting that you can do with AI. I know as a designer like how many times have you been like, "Oh, an illustration would be amazing here, but no one wants to spend the time to like draw a custom icon or a photo would be here, but like we don't have any stock photo budget for this project, so I'm just going to like erase that and put whites space." And so I just think like that piece is so underrated for AI is like getting out of getting us out of MVP mode.异步分享与“激进的多对多分享”
除了同步会议，我们还设立了一个异步分享渠道，大家可以在其中不断分享文章链接和实验成果。例如，产品设计师 Mark 在 Midjourney（AI image generation tool: 一款AI图像生成工具）推出动画或视频功能时，分享了一个在欢迎界面加入可爱动画角色的想法。这在过去需要专业人士或大量业余时间才能完成的工作，现在通过 AI 变得轻而易举，为应用程序带来了更多生机和用户情感连接。
主持人总结道，Brian推动 AI 采纳的两个主要部分是：定期举行的 产品AI 同步会议（我们每两周一次，主持人团队则是每周一次的“AI能量时间”），以及一个异步的分享渠道。主持人特别赞赏Brian幻灯片中提到的“激进的多对多分享”（Radical Many-to-Many Sharing）理念。在 AI转型 过程中，组织常面临信息囤积和“秘密 AI”的问题。这可能是因为员工不确定哪些 AI工具 可以使用，或者担心分享自己的 AI 技能会失去竞争优势。而“公开构建、多对多分享”对于建立健康的 AI转型文化 至关重要，它能消除不确定性，鼓励知识共享，避免个人囤积信息和技能。
Original English
Yeah. And I got actually a really cool example of that as well. So like, you know, that was the sessions, right? But then there's also the channel and so um you know there people are constantly sharing things links to articles experiments that they've tried. I mean that was the whole intent of it right is just to sort of have a space where people can share things and I think there's an example in here where ah yes here it So Mark, you a he's a a product designer and I think this was right around the time that Midjourney um launched the ability to do animations or video and he was like, "Wouldn't it be cool to have, you know, sort of in this intro screen these little animated characters that sit there and just like wave at you, right?" And it's so cute. Exactly. Right. And that it wasn't that hard. I mean, like I think he, you know, he iterated on a couple of problems like this, this is not in the product. It's not in the product yet, but I mean, he was able to create an asset that not that hard to drop in if you have the right spot for it. And like like you said before, it was, oh man, like really cool to have like an illustration or an animation, but I got to go talk to a professional or I got to go spend like nights and weekends working on it. It's like not worth the effort. And now we can bring a little more life into our applications. And that life of course turns into like what I just did, which is like, oh my god, I love it. Which is just like customer connection to your brand, to your product. A little bit of sense of like this team actually really cares about the craft and is going to continue to invest in this product experience all that all the great stuff and then I I just want to call out for folks that maybe missed this in part of the transition. So the two kind of major pieces you put in were these sessions, these product and AI sessions, and you showed us the kickoff deck of what that looked like, both the why and then like let's actually get into it. Let's do it together. And then the second piece, that's the sync piece. And you do those weekly, right? Or just we do them bi-weekly. I mean, we could probably do them weekly, but we've we've done them bi-weekly. Yeah. I um I I I put in a we put in a similar thing. We called them like AI power hours on Friday and it was like every week we we would do that. And then the second thing is the async channel which like if you do not have it you should definitely have it where folks are just sharing and I like this bullet point that you had in your slide that was like radical peer-to-peer share. I forget what it said but it was such a good phrase. Yeah. Radical many to many sharing. And so one of the things that I think organizations often suffer from during this AI transformation is information hoarding and like secret AI. And I think it happens for two reasons. Secret AI can happen because people aren't sure what they can use and we're going to talk about that in a minute. And so they like kind of pretend they're not using AI or they use their like Gmail account, Gmail chat GBT account because they don't want to get in trouble but they're going to use it anyway. And so there's like secrecy because people don't know the golden path of using AI. And then there's like information and skills hoarding right now, which is like the dark side of being an AI agent or an AI change agent, which is people are like, well, I'm the only one that knows how to do this. So I'm going to stand out if I'm like extra good on these things or just get my work done faster or whatever. And so this like build in public many to many sharing is so important for a healthy culture around AI transformation. I cannot emphasize this this one enough.衡量AI采纳度与建立AI知识中心
我们通过多种方式衡量 AI 采纳度。首先，我最初的目标并非是全公司范围的转型，而是专注于产品设计和工程领域。然而，我们的 AI 渠道现在有200多人，远超产品组织人数，这表明 AI 的影响已渗透到其他职能领域。我们每两周举办一次会议，主题涵盖 提示工程（Prompt Engineering: 设计和优化AI模型输入以获得期望输出的技术）、客户反馈分析，以及深入探讨 Google DeepMind（AI research lab: 谷歌旗下的AI研究实验室）的 Gemini（Large Language Model: 谷歌开发的一系列大型语言模型）系列模型等。
此外，我们还设立了 OKR（Objectives and Key Results: 目标与关键成果，一种目标管理框架）来衡量 AI 影响力。我们通过“情绪调查”来了解同事对 AI 的看法，因为人们可能对 AI 抱有各种复杂情绪，从担忧失业到认为它能解决癌症。调查还关注员工是否了解公司的 AI使用政策。许多人私下使用个人 ChatGPT（Large Language Model: 由OpenAI开发的大型语言模型）账户进行工作，不确定哪些数据可以输入。
我们发现，季度初到季度末，所有衡量指标都有所提升，其中对 AI使用政策 和可用工具的认知提升最大。为此，我们创建了一个内部 Confluence 文档，名为“AI知识中心”（AI Knowledge Center: 集中存储AI相关信息和资源的平台），其中包含所有员工需要了解的 AI工具 信息，包括哪些工具获得批准、可共享的数据类型以及如何获取许可。这需要与安全、IT、财务和法务部门紧密合作，确保在快速推进 AI 应用的同时，不危及公司和客户数据安全。主持人强调，建立清晰的 AI使用黄金路径（Golden Path for AI Usage: 组织内部使用AI工具的最佳实践和指导方针）至关重要，它能加速团队的 AI 实验和采纳。
Original English
Yeah. Yeah. So, I think there's like there's several ways I can answer that question. The first is like, you know, I didn't intend to have to implement a companywide transformation, right? Like I I needed to start somewhere or at least I wanted to really focus on my craft and the people that are around me. So that was product design, a little bit of engineering, right? And so product AI is intentionally focused on the area around creating products using AI to do design, product management, engineering, that sort of thing. It wasn't intended to sort of like branch out to you know how does revenue sell better or how does finance do whatever finance does better right? The channel so like it really it was really important for the channel to be public. We have like 200 plus people on the channel now. That is way more than the product organization. So the thing is like even if you're sort of focusing on a functional area there's aspects of that functional area that bleeds outside of the organization. So just to kind of give you a perspective on like the sessions that we ran right so we started back in January we were doing it every two weeks and you can kind of see sort of like some of the the different topics. It wasn't just all about vibe coding. It's about prompting. It's about you know how do you take customer feedback and sort of make sense of it. There's sessions here on just like diving into just Gemini. I mean like we have access to all a whole bunch of Google features from Google that are AI related and they're spread out throughout their entire ecosystem. So it's like hey what's what are all the things that we can do to sort of take advantage of that and that was one where I intentionally made it not just about product and design because like you know the the finance people and the revenue people could definitely take advantage of things like deep research within Gemini or the AI function within within sheets. So sometimes like it makes sense to sort of like really promote these things outside but again it was like you know you kind of want to make sure that you stay within your group. And then separately there was an initiative an OKR in the first quarter of of this year. Our quarters begin in February. So I had started this. Yeah. What's it's so weird enterprise enterprise sales fiscal year right there. Yeah. It it it really confuses me because like we're in fiscal year 2026. Like I thought still 2022 and it's October. What's happening? Yeah, totally. So yeah, we're like a little like one month off. So our quarters begin in in uh in February. So I had started this and I think because you know to your to your point earlier about like how this this is also a good career move. It adds a lot of visibility I think to you if you sort of take the initiative. Like I had just started this a couple weeks earlier, but I was invited to be part of a cross-functional group that was responsible for a companywide OKR to improve AI AI leverage within the organization. And so I can show you some of the things that we did but like I think the most important thing that we did was just measuring right just measuring sort of like what is the and we called it a sentiment survey because like we didn't really know what people's or like my colleagues feelings were about AI right like because you know you might feel like AI is taking your job or AI is creating slob or you might feel that like AI is such like a cool fun like incredibly transformational technology that you know is gonna solve cancer, right? Like I don't know what the the sentiment is. We weren't really, you know, you're not h you were never really hiring for the skill or this attitude, right? And so you have a group of folks and you know you're thinking about instituting a transformation on how they work and the technology they use. And so it was really important I think for us to get a temperature read on like how people felt about AI. But we also wanted to know other dimensions as well. Are they aware of the like for instance are they familiar with our usage policy? Right? There's a lot of shadow IT happening just like you said like people are using their Gmail their personal chat GBT accounts to do professional work and they're not sure is like is that cheating? Is that allowed? Or even like what kind of data am I allowed to put into chat GBT? Can I put like customer transcripts in there? Like I don't I don't know what the answer is. And some people are just doing it because they know it's going to be helpful and some people are not doing it because they're worried that you know that's not that's not cool. And the other thing too is like they don't know what tools are available because yeah, I mean like you have Salesforce, you have Gmail, but you don't like we at the point at that time we didn't have companywide licenses for chat GBT, right? So like no one knows what what is actually available to them. And so what I'm showing here is like the beginning of the quarter. So like the idea was we would do some things within the quarter, but in the beginning of the quarter we take this baseline and we asked these five questions and we also got some some qual feedback as well. So it kind of gave us a little an idea sort of like you know why things were trending in a certain way. And then one thing that we we we noticed and you're kind of looking at sort of like the trend of what happened be from the beginning of the quarter to the end of the quarter. There was an increase along all of these measurements and the biggest ones were around this usage policy and which tools do I have available to me. I think the biggest gap was here because we didn't spend any time with it. So there was a lot of work done in that quarter just by making people aware of what they can do and how they can request software. So here's a just a screenshot of an internal Confluence document we call the AI knowledge center. And in this document is all the information that an employee needs to know about which AI tools they have available to them. So, like if you were to scroll down, you'd get this like alphabetized table of all the products that have been approved to be used within our organization for security reasons, for for legal reasons. I mean, the the thing is that like AI is a is a vector for doing some really bad stuff. And even though you want to move fast and you want to use all of these really cool tools, you don't want to put your company and your customers data at risk. And so it's really important that you know you work closely with your security, your IT department, your finance department, your legal department. And like again I was very fortunate in to be in an organization where like those folks which sometimes can feel like friction and a barrier like they I think they recognize that like this was also really important and sort of like prioritized you know still doing all the the solid work but like adding I prioritize sort of like the enabling us to sort of like not just use these tools but experiment with different ones. like I found myself for like a month like every week I was submitting these like zip requests for new software that I wanted to to try out. And it would only take maybe like a week for me to sort of get the the okay and you can kind of see like it was really important not just to see which applications I had access to and how how that application could be used but what kind of data can I share within an application as well as if I want to get access to it like what are ways that I can I can get to that. And so when I you know I go back to this slide you know that was that was one of the the the tactical things we're able to get done in a single quarter that really made a huge difference sort of in this um in this metric of awareness about policy and tools.内部价值展示：MCP服务器的突破
主持人总结道，Brian通过成为变革推动者，启动组织对 AI 的重视，利用同步会议和异步 Slack 渠道推动持续实践，并通过 OKR 衡量实际效果，成功提升了员工对 AI 的积极情绪。
在过去一年中，Brian最引以为傲的 AI项目 是为 Pendo 构建了自己的 MCP服务器（Model Context Protocol Server: 一种用于管理和协调AI模型上下文的服务器）。他观察到，像 Deep Research 这样的 AI代理（AI Agent: 能够自主感知环境、决策并执行任务的AI系统）可以反复运行网络查询来构建研究文档。在增长工作中，产品经理经常需要监控转化率和留存率仪表盘，当数据出现异常时，往往需要花费大量时间进行迭代研究。Brian认为，如果能让 AI工具 访问公司内部数据（如使用数据），并利用其智能分析数据，例如按区域或年份进行比较，将极大地提高效率。
MCP 是一种能够实现这一目标的技术。尽管 Pendo 使用的是非 SQL 的专有查询语言，但Brian找到了克服这一障碍的方法。他利用 Cursor 和 MCP 文档，在几个晚上内就构建了一个 MCP服务器。虽然他对服务器的内部工作原理并不完全了解，但他成功地展示了其价值。他通过 Claude（Large Language Model: Anthropic开发的大型语言模型）向 MCP服务器 发送查询，服务器返回了数据，并生成了一个展示过去30天最常用页面、功能和指南的仪表盘。这种将 Pendo 数据与 AI 结合，通过简单输入即可生成可视化仪表盘的能力，彻底改变了人们的看法。
Original English
I want to just call this out for the leaders on the team that are or the leaders in the audience that are listening. This is the first thing I tell them to do is I say define the golden path to using AI and it takes three pieces. It takes finance and procurement. It takes legal and it takes security. And what I tell them is it's really not going to benefit the acceleration of your team to say we'll go heads down and figure out how you can get chat GBT and Cursor and you'll get your three little little tools and we'll let you know what they are. You actually need what you called out which is a very fast path to experimenting with reasonable tools to identify which ones are going to work for your team. And that rapid experimentation is really really important. You can't go do a big like multi-month evaluation of one code editing tool because as you said they're changing so rapidly. And so I love this documented place of like here's the tool, here's the status, here's the data you can put in and the data you can't put in. Here's how you get a license. Here's how you get help. Is just very very very useful. And and if you can get this done, then it all starts to snowball from there because people have a place and a path to go down. So, I this is probably like not the most exciting screen share we've seen. Like, it's a it's a table and a confluence talk, but like I just want everybody to pause, screenshot this if you do not have this in your org. Like, you need it. Need it today because this is going to be the thing that changes how you work. I love this. Yeah, 100%. And like once you kind of get that ball rolling, like you know, there there's a separate channel too, like I think it's called like the AI knowledge center. It's meant to be sort of like the product AI channel but more like broad-based. And every once in a while, someone be like, "Hey, can I use granola for this?" And someone outside of it, you know, someone who's just familiar with this with this process would just like, "Hey, yeah, go check out the thing. You know, yeah, you can use it, but you have to get approval, yada yada yada." And so like once you sort of get the ball rolling, you have really good documentation, you just sort of reemphasize this is where you go, this is the process. And on the on the flip side, I mean, I'm not in legal, security, IT, that sort of thing. But like when that those groups that that group those groups are responsive to an organization that wants to experiment and try this stuff, then that the the flywheel just just keeps going. And you know like I said in the beginning of the year there was a lot of like confusion about what I can do and whatever and now I wouldn't say it's all gone but like it is nowhere near the top list the top of our list of things that like we are concerned about when it comes to AI transformation. Well, I love this. You know, just to recap, you've shown us how you become a change agent. You know, sort of incept your organization into taking this seriously as an initiative, how you use synchronous meetings and asynchronous Slack channels to drive this as a consistent practice over time. And then you use OKRs to actually measure does any of this matter? And you're showing that like if you put these simple things in place, you can actually inlect those measures which you know just in the looking at the sentiment. I like that last question that you had on the sentiment survey which is like I think this is going to have a positive impact on our employees and that went up and that is a huge win for for a company because a lot of people are feeling fear about their careers, uncertainty, doubt and so the fact that you can show you can do these very simple things and inflect that sentiment very positively in your employee population is also fabulous. So, Brian, this was great. Thank you for showing us again like I could not write a better playbook for getting AI adoption in a team. This is what I've done. I you it works. Let's do this. Let's repeat this and stamp it out in our our own teams. So let's get to some lightning round questions and then we'll get you back to all your fun AI projects. The first question I have is okay. You showed us like you're incepting people and you're you know managing your stakeholders and all that kind of stuff. What is your favorite thing you've built with AI over the last year for for work? I've built a lot of really cool things. The one that I'm most proud of, but it's it's kind of geeky is that I built my own MCP server for Pendo. So, back maybe back in the spring whenever when MCP was becoming really hot, folks have been talking about MCP and like how it could be applied to to Pendo. And what I clearly saw was after using tools like deep research that is essentially like an agent that can you know basically run a bunch of web queries over and over and over again to sort of build up sort of a a research document. I had spent you know I've been in in tech for a while. I I'm a designer now but I've also spent time as a PM and I've done a lot of growth work. And one of the things that is common within growth is you're you're constantly monitoring like dashboards of conversion rates and retention and blah blah blah. And every once in a while a number will like dip or will dive, right? And then you're you're off and you're spending like the entire week trying to understand like what is it like? Do we have like an issue? Do we have a bug? Like is there something going on, you know, in one of our regions? You know like if you're doing a year-over-year comparison like maybe a holiday fell you know last year and not this year and so like that's why these and usually it's it's like some common sense thing but like if someone above you sort of sees this and they're just like wait a minute like fire drill like we got to go figure it all out and that like just eats up your whole week doing like a lot of like iterative research right looking at the data from different angles and I'm like this could be like like a deep research like query but having access to like our data right um or usage data could be uh like a killer and like NCP is a technology that can leverage that if I can give the AI tools to run these queries and then like use its intelligence to like look at the data and be like oh what about looking at it by region let's look at it a year let's look at it blah blah blah um and so that was sort of like the idea that I had of connecting like okay how this is how MCP could be useful for our customers right because we also we're a product analytics um provider and so I had this conversation with my colleague colleagues and they're just like I don't know if you know because we have like a a bespoke quering language. It's not SQL, so it it takes, you know, the AI doesn't know how to write those queries, but I I thought I had a way of doing it that that could sort of overcome that. And so over a couple of nights, I basically just leveraged the MCP documentation giving it to Cursor. I have no idea how my I built an MTP server. I have no idea actually how it works, but I understand like right. It's like a lot of I'm laughing. I'm laughing. Everybody close your ears. I legitimately approached MCP servers as true sorcery sorcery before I built one. I was so I don't there's there's some branding issue that that the MCP platform and framework has because it seems like sorcery. So I was with you until I actually built one and like you know wrote the code and then I was like oh okay I get it a little bit. There's a lot of elements that make it like really hard to gra like model context protocol. Like what the hell do those words mean? Yeah. Like it doesn't actually speak. Even the naming it's just okay. So MC framework caretakers you might as well call it magic context protocol for for all we know. Yeah. Yeah. What one of my most popular sessions like so you know it's not just me doing the sessions during product day. Other people will do them, but like one of my most popular one one was called WTF is MCP and I just spent like a whole hour telling people what MCP was and afterwards they're like, "Oh, I get it now. It's kind of like web search but like more tools." I'm like, "Yeah." But anyway, so I I I built this thing and like I have like so you know my wife also so works in technology but she's in marketing and I'm sitting like next to her and you know how it is like you have a baby and like you know your your nights are a little bit slower and so you have some slack time just kind of like hang out and she's watching TV and I'm sitting there on the laptop like working on this MCT server and I got it to do a thing right I got it to like you know I wrote a query a prompt within Claude and it talked to my MCP server and it ran a thing and it returned data back and I was just like, "You have no idea how how credible this is." And she's like, "I don't know what an MCU is." Like, "What is an MCU?" And I'm like, "It's not MCU, it's MCP. God damn it." So I recorded so I recorded a demo of me using it. And this isn't the exact one, but this is something very similar, which was um, I so I have a the my own MCP server that I hooked up and I was I was just using the public APIs, right? So like I have test accounts for Pendo. I didn't have any special access. I don't have a dev platform. I was just able to like do this on top of the APIs. Actually, any customer could do this. And so I had this MTP server that can do things like just grab really basic stuff like how many pages which pages are the top pages, how often they accessed, you know, what are things, you know, what what does this this visitor do and that visitor do? Um, so I was like, can you create a beautiful dashboard that shows me which pages, features, and guides were the most used over the last 30 days? And I think the key thing was that I'm combining accessing the data and having it create a dashboard. So I won't kind of go through all the things I did made a bunch of tool calls. That's not really important. But the important thing was at the end I had it put the data into this artifact. I'm I'm sitting here in Claude and it's not I mean I wouldn't ship this but just visualizing the connection between taking Pendo data which now you know only exists within Pendo or you can also export it and there's other ways of getting access to it but making it accessible on the level where someone could just type in a thing like tell me who what the top pages are and then show me a dashboard really like changed people's minds and with over the next like couple months like I'm now chatting with the CTO. The CTO is creating another versions MCP and you know he's so he's create like we're we're iterating creating starting to like productional productionize some of these elements. I mean what I've done here like they just they kind of moved on the it wasn't the code that I wrote or vibe coded was not important. What was what was important was like I was able to demonstrate the value of this somewhat you know hard opaque technology to folks that like didn't maybe didn't see it in the same way and that has now actually that has like significantly impacted our road map. So like we're working on a lot of agents within Pendo that are leveraging sort of the MCP but you know behind the scenes and there's other things that we're planning to do with MCP and we might have gotten there anyway. But that that definitely sort of accelerated our timeline for that.掌握硬技能，影响产品路线图
主持人祝贺Brian取得了突破，并强调他所说的“你写的代码不重要”是完全正确的。关键在于你能够向组织内部展示价值。对于像 AI 这样抽象的技术，只有当人们能够“触摸”到你的想法时，它才能真正被理解和采纳。通过使用 AI工具 让同事能够亲身体验你的想法，你就能更好地影响产品路线图，这正是许多设计师、工程师和产品经理所渴望的。
Brian补充道，对于设计师和产品经理来说，理解 AI技术 的工作原理至关重要。你不需要成为 机器学习工程师（Machine Learning Engineer: 专注于设计、构建和部署机器学习模型的工程师），也不需要编写复杂的代码，但你需要理解 大语言模型（Large Language Model, LLM: 基于海量文本训练的AI系统）和 AI代理 的基本概念。就像建筑师不仅设计美观的建筑，还要了解水暖和电力系统一样，产品人员也需要掌握这些技术底层知识，才能将可用技术与客户问题连接起来。主持人对此深表认同，并强调这是“硬技能”的时代。他指出，Brian的某个会议主题是“HTML和CSS入门”，这正是解锁设计师利用 AI 编码工具的关键。
Original English
Well, congratulations for for cracking through that. And I think what you said is exactly right, which is like the code that you wrote is not important. Like people get so wrapped around the axle on the quality of like quote unquote vibecoded stuff. Not the issue. The issue is until you can internally display value, it is really hard to get anything done in a product organization. And sometimes you can internally display value by saying a customer wants this. Like very easy sales comes. I got like a top quarter making deal. Customer wants that. Okay. Like we love it. But something that's a little bit more nebulous, something that has a little bit more of an intangible value. Like I think we could use this new technology or this new framework MCPS to give our customers a better experience of our product. Like it just doesn't click until you can touch it. And so the more you can use these tools to let your peers sort of like touch your ideas, I think the more as you said you can like impact the road map which is what I hear all the time from designers, engineers, product managers like I'm tired of the road map being given to me. How do I like impact it? And this tool these these tools are definitely a way way of doing that. Yeah. And I would say one just one tip around this which I think is really important to emphasize is that you know you might sit sort of like you know you might be designer or PM I think it's really important for everyone to understand how this technology works like I think you got to get a little more technical you got to understand how LLMs are working you have to understand what like what an agent is you don't have to be able to code these things you don't have to be an ML engineer but if you are a creative person or someone who is sort of in the solution space and is trying to think about like what like what are the different ways that I can apply AI to this this problem. You won't arrive at anything really interesting unless you understand the underlying technology. It's like an architect. Like architects, yeah, they they design like pretty buildings, but they also understand how plumbing works and electric and the electronics work, right? Like they're they're not like the electricians and the plumbers, right? But like in order to build a functional building that stands up and and is functional, right? like you got to have power sockets, you got to have room for the pipes, like all that stuff is important. I think the same thing applies to PMs and designers. And I think if you're an engineer, I think it's also useful to understand like the business side, the customer side, the you know, like to really empathize with the the issue, the the problems that your customers have so that it, you know, like what we're all looking for is basically how do you sort of connect the dots between the technologies that we have available to us to solve customer problems. H I couldn't hype you up more because I have been screaming from the rooftops. This is the era of the hard skill. Like you need the hard skills to take advantage of this stuff. I I didn't call it out earlier, but I saw one of your sessions was like intro to HTML and CSS. Like that is actually going to be the unlock for your designers because if you can read CSS, a whole world has opened up to you with these vibe coding tools. And so yes, you need to know how the stuff actually actually works.AI不听话时的应对策略
当 AI 不听话，或者 MCP服务器 无法正确进行“Vibe Coding”（Vibe Coding: 一种非正式的、基于直觉和实验的AI代码生成方式）时，Brian的应对策略是：先“吼一吼”（有时会），然后会说：“好吧，你没完全理解，换个思路试试。”他会尝试引导 AI 跳出固有的思维模式，进行更广泛的思考。如果吼叫无效，他会要求 AI 尝试五种不同的解决方案。主持人认为这比他自己“不”一声更有效。
最后，Brian鼓励对产品构建或在拥抱 AI 的公司工作感兴趣的人，可以关注他的 LinkedIn（Professional social network: 一个职业社交网络平台）或查看 Pendo 的机会。
Original English
Okay, so Brian, last question. This has been so great, but got to have your your strategy here. When AI is not listening, when that NCP is not being vibe coded correctly by Cursor, what is your go-to tactic? How do you get it unblocked? Do you yell? Uh, you know, I I I don't want to say that I don't yell, but I do yell sometimes. The the other thing that I do too, which I think is also really helpful, is I say, "Okay, you're not you're not quite getting it." Like think think about a different way of approaching this, right? Like I I'm trying to nudge it to like it maybe it's sort of like locked into a certain sort of uh groove and I'm trying to to make it think a little more broadly. I I don't know if that actually helps, but I've found that like sometimes that's useful. So like if I'm not yelling, it's just like, "Okay, you tried this a million times. Think about five other different ways that you can solve this problem." and and and go for it. Yeah, I think that's probably more effective than my no that I do when it's not working. Well, Ryan, this has been awesome. Where can we find you and how can we be helpful? Yeah, so you can find me on on LinkedIn. I try to post when I can and um yeah, I mean, if you're if you're a product builder or um you know, you're interested in working for a a company that is sort of embracing AI, you should check out Pendo. Awesome. And thanks so much. Thanks so much for watching. If you enjoyed this show, please like and subscribe here on YouTube or even better, leave us a comment with your thoughts. You can also find this podcast on Apple Podcasts, Spotify, or your favorite podcast app. Please consider leaving us a rating and review, which will help others find the show. You can see all our episodes and learn more about the show at howiaipod.com. See you next time.📌 文中提及的人物和组织
人物: Mark
公司/组织: Google, Google DeepMind
产品/模型: Cursor, Figma, Midjourney, Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.5 Flash Light, Claude