请问HN：当你的界面是聊天时，您如何理解用户的需求？
构建一个对话式人工智能产品时，遇到了一个具体问题：传统的分析方法（点击量、漏斗、热图）在没有用户界面的情况下毫无用处。<p>想知道其他创始人是如何处理这个问题的，你是如何了解用户实际想做什么的？哪些意图最常出现？他们在哪些地方感到沮丧？<p>你是在手动阅读对话吗？在构建内部工具吗？还是在盲目操作？
查看原文
Building a conversational AI product and struggling with a specific problem: traditional analytics (clicks, funnels, heatmaps) are useless when there's no UI to track.<p>Curious how other founders handle this, how do you know what your users are actually trying to do? What intents are coming up most? Where are they getting frustrated?<p>Are you reading conversations manually? Building something internal? Just flying blind?