修复状态更新——我们的宣言
1 分•作者: xvok•大约 1 个月前
我们解决的问题是组织摩擦和可见性问题。随着层级的增加和会议的增多，你离真正了解情况会越来越远。我们构建了一个工具，让任何人只需点击一下按钮，就能看到项目实际的进展情况。无需与人工智能争论，无需解读甘特图。只需一个合理的、人类能理解的解释。
当然，我们一开始不会做到完美。我们专注于做好一件事，让项目经理或总监点击“更新”，就能以他们期望的方式获得更新。这意味着我们需要“了解”他们的项目。为此，我们正在使用人工智能。
我们不期望用户在我们的工具中启动项目，也不期望他们调整现有的工作流程。我们消化一切，然后输出状态更新。这意味着我们的工具在结构上能够像项目经理一样理解事物。有些人可能会说，这意味着我们的工具无法完美契合他们的项目或计划。对此，我想说：“你并没有你想象的那么有独创性。”这意味着人工智能从一开始就具备规则和技能，能够将事物分类并进行自我调整。我们一开始并不花哨，我们提供了 Jira 集成和文档上传功能。
现在，还有调整的空间。通常情况下，这些调整是在人工智能聊天中进行的，而对我们来说，这些调整体现在规则和技能中，以及结构化的提示。但是，用户并不关心这些。相反，当状态更新返回时，他们会进行一两次调整。然后，我们会在后端调整规则，以便下次应用。所以，我们一开始的准确率是 80%，但每次用户使用后都会变得更好。而且，嘿，对于你的项目、你那些不规范的格式和你的语气，80% 的准确率已经相当不错了。
那么，最终结果是什么？对最终用户来说，这意味着他们可以登录工具，点击“运行”，看到它运转片刻，然后获得一个他们可以轻松使用的状态更新。我们提供的是一种“一键生成状态更新”的体验，取代了手动整理文件和项目以获取状态更新，或者更现代化的方式——与 Claude 的输出进行协调、恳求和最终妥协。
简而言之，想象一下这样一个世界：当有人问“有什么更新吗？”，你只需按一下按钮，他们就会自行离开，或者给你所需的信息以继续工作。就是这样。
这里的愿景是什么？我的意思是，如果你在大公司工作过，我们肯定认为在削减状态更新报告中的水分方面，有很大的商机。从这个角度来看，我们的整个愿景是解决公司里人们花费时间最多的项目管理环节。这首先包括状态更新，但将来还会涉及更多内容（例如项目章程、RACI 等）。
想试用我们的 Alpha 版本吗？请通过 logan@quickapproveai.com 或 uzma@quickapproveai.com 与我们联系。
查看原文
Our problem we solve is the issue of organizational friction and visibility. With each layer and each meeting, you get further from really knowing what is going on. We build a tool that lets anyone at the click of a button see what is actually going on with a project. No arguing with an AI. No divining with a gantt chart. Just a reasonable human explanation.<p>Of course, we won’t hit perfection right out the gate. We’re focused on doing one thing right, letting a single program manager or director hit “update” and getting an update in the way they expect. This means we need to “know” their project. To do that, we’re using AI.<p>We don’t expect users to start projects in our tool or to adjust anything they do. We digest everything and burp up status updates. This means we structure our tool to understand things like a program manager would. Some might say that means we won’t fit their projects or programs perfectly. To that, I say, “you’re not nearly as original as you think you are.” This means AI rules and skills out the gate that self adjust as it sorts things into categories. We aren’t fancy out the gate, we have a Jira integration and a document upload.<p>Now, there are room for tweaks. Where these normally live in an AI chat, these live in rules and skills for us along with structured prompts. But, users don’t care about that. Instead, as status updates come back, they tweak things once or twice. We then adjust the rules on the backend so they get applied next time. So, we start at 80% accuracy, but we get better every single time they use it. And, hey, 80% right on your projects,your janky format, and your tone is pretty good.<p>So, what’s the end result? To the end user, it means popping on to the tool and hitting “run” and seeing it whirl for a moment or two before they get a status update they can jam where they need it. Taking things from either manually sorting files and projects to get status update or the more modern version of corralling, begging, and finally compromising with Claude output to a simple “one-click, one status update” we’re offering.<p>In short, imagine a world where when someone asks “any updates,” you hit a button and they go away and/or give you what you need to continue. That’s it.<p>What’s the vision here? I mean, if you’ve been at a big company, we certainly think there’s a killing to be made in the art of cutting out the baloney in producing status updates. To that degree, our whole vision is to attack the project management pieces taking the most time for people in the company. That means status updates at the beginning, but many more things later on (looking at you project charters, RACI, and such).<p>Want to try our alpha? Reach out to us at logan@quickapproveai.com or uzma@quickapproveai.com