伪需求错觉与真实产品契合度
在创业初期,创始人往往容易陷入“伪产品契合度”的陷阱。我曾在杜克大学的神经科学实验室工作,随后辍学创办了第一家公司,试图利用脑部扫描来测试广告效果。直到某天,一位客户问我:“我能把这个头环带回家吗?我想戴在我儿子头上,看看他写作业时什么时候在集中注意力,因为他患有多动症(ADHD)。”这个真实的故事让我改变了方向,公司转型为研发针对多动症儿童的消费级硬件,并更名为 Neuroplus。在这7年的运营中,我们发起了成功的 Kickstarter 众筹,甚至让我登上了福布斯30岁以下精英榜。尽管我们拥有一批极其狂热的粉丝,这却让我产生了一种错觉,以为我们已经找到了真正的产品契合度,或者离爆发式增长只有一步之遥。然而,一两个优秀的客户访谈并不能代表市场的全貌。如果你的早期种子用户群体只是一小拨极其狂热的早期采用者,且与大众市场之间不存在自然的过渡梯度,那么你的市场空间实际上是微乎其微的。
真正的产品契合度(Product-Market Fit: 产品真正满足了广阔市场的强劲需求),是我后来加入 Verkada 时才切身感受到的:在第一次与客户通话时,虽然销售人员在产品演示中表现糟糕、频频失误,但在电话结尾,客户依然平静地说:“听起来不错,能给我发个报价吗?下个月我们大概会买30台摄像机。”当你找对方向时,你不需要在每一个商业环节都做到完美无缺,市场本身那近乎蛮横的拉力会自动解决诸多难题。
Original English
I was working in a neuroscience lab at Duke... I dropped out of college to start my first company at a company testing advertising with brain scans... one of our customers said, "Hey, can I take that headset home? Can I put that on my son and see when he's paying attention to his homework and when he's not? He has ADHD..." And we ended up pivoting the company to build out a consumer product for kids with ADHD, building new hardware, a new headset, and that became Neuroplus. ran that company for about seven years. We had rabid fans of our product that loved what we were doing. And that gave me this delusion that we had found product market fit or we were just around the corner from really unlocking massive growth... we just done a successful Kickstarter campaign... But so at the time when I got Forbes 30 under 30, I didn't think of the company as falling apart... So maybe we did have some product market fit, but it was with a market that was so small that it it was pretty irrelevant and inconsequential... It's only in retrospect when I joined Vicata and I saw real product market fit that you see how different it is. So, when I was on first call with a customer, a sales rep was demoing the product and I didn't think the sales rep did a very good job and they messed up a lot... And then at the end of the call, the customer says, "Okay, sounds good. Could you send me a quote? I think we'll probably buy 30 cameras or so in the next month."... you don't have to be perfect at every other aspect of the business if you get that part right.
平台级产品的定力与客户深潜
在建立这种对真实痛点的认知后,我们创办了 Serval,一个专为 IT 团队打造的 AI 原生自动化平台。我们意识到,它不仅是一个处理密码重置或 Wi-Fi 查询的简单帮助台,而必须是一个全面融合 ITSM、访问管理和工作流引擎的企业级枢纽。在这个前提下,单一功能的修补绝不会带来爆发,我们必须忍受长达一年的建设期,在毫无市场牵引力的情况下坚定信念。直到2025年4月或5月,当我们终于完成闭环并签下最初的几单后,奇迹发生了:客户沟通的基调几乎在几周内发生了翻天覆地的变化,从客套的“这很棒,保持联系”瞬间转变为“这要多少钱?我们的采购流程是怎样的?”。
在这种关键的成长期,传统的客户发现(Customer Discovery: 深入理解客户痛点与使用场景的系统过程)机制必须被重构。在早期的职业生涯中,每次与客户交谈后我的情绪就像钟摆一样大起大落。但现在,我对单一的正面或负面反馈保持着绝对的情绪中立。取而代之的是,我每天花5到6个小时沉浸在客户对话中,甚至直接潜伏在他们的 Slack 频道里。我不再将这看作是生硬的“客户访谈”,而是与他们建立长期的伙伴关系。只有当你出于纯粹的兴趣去解决他们的麻烦,全面嵌入他们的日常工作时,你才能培养出比死板问卷强大百倍的系统直觉,从而打造出真正令市场惊艳的产品。
Original English
Servil is an AI native platform for IT teams. We help automate help desk requests, onboarding, offboarding, just in time access requests... we had a theory that the product that we wanted to build had to be a platform. It had to combine the full ITSM this AI native workflow builder access management help desk automation... we wouldn't really have a product to test until we built the platform. And so that's how we kind of kept the conviction during that one-year period... the tenor of the conversation changed almost overnight over a few week period in April or May of 2025 where it went from this is really neat keep us posted to how much does it cost how do we get started... So, you become very neutral to the customer feedback. Whereas, I think in the early part of my career, every customer conversation, you know, you just you're a pendulum... I want to be embedded with customers. So, I don't think about customer interviews or customer conversations. Like, I'm just talking to customers all the time. I'm in their Slack channels... That builds just an intuition for our customers that is much much more powerful than these kind of point in time customer interviews... It has to be a genuine interest in them and in making their lives better that drives that relationship forward.
AI产品的技术前瞻与战略博弈
对于如今的开发者而言,构建 AI 产品的最大技术博弈在于:你必须时刻预判底层大语言模型(Large Language Model: 基于海量文本训练的 AI 系统)的进化轨迹。在 Serval 刚起步时,AI 的能力尚不完善,工作流引擎无法仅凭一句话就顺滑生成所有操作,我们不得不在后台投入大量的提示词工程与复杂机制来弥合体验差距。
这引申出了一个极其艰难的核心抉择:你是基于当前能力平庸的模型来规划产品,还是押注于未来即将大幅跃升的技术奇点?我们的策略是“滑向冰球将要到达的地方”。我们坚信,通过我们在工程架构底层的巧妙设计,加上大模型必然的能力提升,最终将兑现全面自动化的愿景。然而,绝对不能把“技术前瞻”异化为逃避现实的借口。有些创业者会自欺欺人地认为:“虽然现在完全行不通,但只要模型进化了,我们自然就能成为一家伟大的公司。”这是极其危险的幻想。最明智的战略,是果断拥抱那些几乎可能(Almost Possible)的硬核挑战,只需一点点信念的跨越,你就能在技术彻底成熟的前夕,精准抢占属于你的生态位。
Original English
I think when you're building an AI product today, one of the most challenging things to always be thinking about is what's going to happen with the models and how are these capabilities going to change over time... the workflow builder did not allow you to just describe any arbitrary automation and have that build out automatically without intervention. It took a iteration. It took um prompting. It took a lot of behind-the-scenes minations to make that all work... And I think the bet that we took was that between our efforts and improving the product and doing a lot of clever things under the hood and the improvements in the underlying models that we would be able to build a product that delivered on this vision... Are you building for how the models work and behave today or are you building for a future where they're a little bit better or are you building for a future where they're a lot better... And on the one hand, you want to kind of be skating ahead a little bit and knowing, hey, like we did where it's like the models are not good enough for what we want to do yet, but they will be very soon. In the same way, you can delude yourself into thinking, oh, this doesn't work today, but all we need is the models to get better and then all of a sudden we'll have a company. I think that's a bad idea. But I think it is smart to embrace challenges that are almost possible and you could with a small leap of faith believe that they're going to be possible in the near future.