How Fyxer built an AI executive assistant people trust
How Fyxer built an AI executive assistant people trust
Fyxer 如何打造出一款备受用户信赖的 AI 行政助理
Fyxer pairs OpenAI models with 500,000+ hours of EA workflows and real user feedback to draft replies in each person’s voice. Fyxer 将 OpenAI 模型与超过 50 万小时的行政助理（EA）工作流程及真实用户反馈相结合，能够以每位用户独特的口吻起草回复。
For many professionals, work means keeping track of conversations and commitments across inboxes, meetings, messages, and apps. Without rock-solid context, those commitments can fall through the cracks, damaging projects and relationships. 对于许多专业人士而言，工作意味着要跟踪收件箱、会议、消息和应用程序中的各种对话与承诺。如果没有扎实的背景信息，这些承诺可能会被遗漏，从而损害项目进展和人际关系。
Fyxer built an AI executive assistant that follows the thread as work moves across tools. It combines the latest OpenAI models with more than 500,000 hours of executive assistant workflows, dividing the work among dozens of specialized models that improve through real user feedback. Fyxer 开发了一款 AI 行政助理，能够随着工作在不同工具间的流转而持续追踪任务脉络。它将最新的 OpenAI 模型与超过 50 万小时的行政助理工作流程相结合，并将任务分配给数十个专门的模型，这些模型通过真实的用户反馈不断优化。
Email is one of the clearest places to see it in action. Two people can receive the same email and need completely different replies depending on the relationship, what has happened before, and what each person is trying to get done. That makes a seemingly simple task deceptively hard for AI. 电子邮件是体现其效用的最直观场景之一。两个人收到同一封邮件，却可能需要完全不同的回复，这取决于双方的关系、过往的沟通记录以及各自想要达成的目标。这使得看似简单的任务对 AI 而言却极具挑战性。
“There’s something called Moravec’s paradox,” explains Fyxer Co-founder Archie Hollingsworth. “Things that humans find easy are hard for computers, and things that computers find easy are hard for humans.” Fyxer handles that complexity by learning how each user works, so it can respond like an assistant who already knows what matters. “有一种现象叫莫拉维克悖论（Moravec’s paradox），”Fyxer 联合创始人 Archie Hollingsworth 解释道，“人类觉得简单的事情，计算机做起来很难；而计算机觉得简单的事情，人类做起来却很难。”Fyxer 通过学习每位用户的工作方式来处理这种复杂性，从而能够像一位深知工作重点的助理那样进行回复。
Frontier OpenAI models support key parts of the experience, from understanding the email and finding the right context to generating the draft. Fyxer chose OpenAI because its models performed best on Fyxer’s internal benchmarks, offered strong fine-tuning capabilities for subjective tasks like tone and intent, and provided hands-on engineering support through whiteboarding sessions and technical collaboration. OpenAI 的前沿模型支撑了该体验的关键环节，从理解邮件、寻找合适的背景信息，到生成草稿。Fyxer 选择 OpenAI 是因为其模型在 Fyxer 的内部基准测试中表现最佳，为语气和意图等主观任务提供了强大的微调能力，并通过白板会议和技术协作提供了实操性的工程支持。
“We chose OpenAI because they have the best models, and they’ve given us real access and a close working relationship. I can drop a question in Slack and get an answer quickly, and when we face a problem, the team comes to our office and works through it with us. They show up.”—Archie Hollingsworth, Co-founder, Fyxer “我们选择 OpenAI 是因为他们拥有最好的模型，并且给予了我们真正的访问权限和紧密的合作关系。我可以在 Slack 上提问并迅速得到解答，当我们遇到问题时，他们的团队会来到我们的办公室与我们共同解决。他们说到做到。”——Archie Hollingsworth，Fyxer 联合创始人
Fyxer’s approach offers three lessons for founders building highly contextual AI products: Fyxer 的方法为那些构建高度依赖背景信息的 AI 产品的创始人提供了三点经验：
1. Break email into smaller jobs
1. 将电子邮件拆解为更小的任务
Fyxer built its system around 30–50 specialized models, each responsible for a narrow part of the email workflow. Instead of treating email as a single text-generation task, Fyxer breaks the problem into a system of predictions, like deciding whether a message requires a reply or drafting responses that match a user’s tone and context. Fyxer 的系统围绕 30 到 50 个专门模型构建，每个模型负责电子邮件工作流程中的一小部分。Fyxer 没有将电子邮件视为单一的文本生成任务，而是将问题拆解为一个预测系统，例如判断某条消息是否需要回复，或者起草符合用户语气和背景的回复。
“Breaking the problem into many smaller models works much better than asking one model to write a good email,” Hollingsworth explains. “将问题拆解为许多更小的模型，效果远好于要求单一模型去写出一封好邮件，”Hollingsworth 解释道。
When a new email arrives, a reply decision model classifies the message: is this something that needs a response, a scheduling action, or simply information the user should see? 当新邮件到达时，回复决策模型会对消息进行分类：这是需要回复的内容、需要安排日程的操作，还是仅仅是用户应该知晓的信息？
If a response is needed, additional models analyze the intent of the email and predict the likely outcome of the interaction. These models determine patterns such as whether the conversation is moving toward scheduling a meeting, resolving a request, or continuing a longer relationship thread. 如果需要回复，其他模型会分析邮件意图并预测交互的可能结果。这些模型会识别模式，例如对话是趋向于安排会议、解决请求，还是延续更长期的关系脉络。
Memory is one of the most important parts of the system. Fyxer needs to decide which details should persist across conversations and which should disappear after a single exchange. When a new email arrives, retrieval models compare it with stored interactions and surface the memories most relevant to that person and thread. 记忆是该系统最重要的部分之一。Fyxer 需要决定哪些细节应该在对话中持续存在，哪些应该在单次交流后消失。当新邮件到达时，检索模型会将其与存储的交互记录进行比较，并提取出与该人员和对话脉络最相关的记忆。
OpenAI models power steps across Fyxer’s system. “We use OpenAI models for everything from digesting the email, so we can understand what it is actually about, to pulling in and re-ranking the context we want to include, to the actual email generation,” Shantsila says. OpenAI 模型驱动了 Fyxer 系统中的各个步骤。Shantsila 表示：“我们使用 OpenAI 模型处理一切事务，从消化邮件以理解其核心内容，到提取并重新排序我们想要包含的背景信息，再到最终的邮件生成。”
2. Train on how great assistants actually work
2. 基于优秀助理的实际工作方式进行训练
Before launching its AI product, Fyxer spent years operating a human-powered executive assistant service. Over time, the team accumulated a dataset built from more than 500,000 hours of annotated executive workflows, capturing how real assistants manage professional communication. 在推出 AI 产品之前，Fyxer 花了数年时间运营人工行政助理服务。随着时间的推移，团队积累了一个数据集，其中包含超过 50 万小时的标注行政工作流程，记录了真正的助理是如何管理专业沟通的。
Those examples gave Fyxer training data drawn from the job itself. They captured the small judgments behind a good response: when to answer quickly, when to wait, which earlier conversation matters, and how the same request can call for a different response from one person to another. 这些案例为 Fyxer 提供了源自实际工作的训练数据。它们捕捉到了优秀回复背后的细微判断：何时快速回复、何时等待、哪些过往对话至关重要，以及同一个请求如何针对不同的人需要不同的回复方式。
Fyxer uses supervised fine-tuning and Low-Rank Adaptation (LoRA) across its broader system to create task-specific model variants while controlling training cost. Early in the product’s development, the team used OpenAI’s fine-tuning platform for tasks that needed high accuracy. More recently, Fyxer worked with OpenAI’s managed fine-tuning team to put a new checkpoint into production. Fyxer 在其更广泛的系统中使用了监督微调和低秩适应（LoRA）技术，在控制训练成本的同时创建特定任务的模型变体。在产品开发的早期，团队使用 OpenAI 的微调平台来处理需要高精度的任务。最近，Fyxer 与 OpenAI 的托管微调团队合作，将一个新的检查点（checkpoint）投入生产。
“OpenAI has been pivotal for us in helping us transfer the learning that we have about our customer and successfully incorporate it into how the models work.”—Joey Dwonczyk, AI/ML Product Engineer, Fyxer “OpenAI 对我们至关重要，他们帮助我们将对客户的了解转化为模型，并成功将其融入到模型的工作方式中。”——Joey Dwonczyk，Fyxer AI/ML 产品工程师
Before any model is deployed, Fyxer evaluates it on validation sets built around its own email tasks, including drafting, classification, and prioritization. The team weighs accuracy alongside response time and cost, since the best choice can vary by job. 在任何模型部署之前，Fyxer 都会在围绕其自身电子邮件任务（包括起草、分类和优先级排序）构建的验证集上对其进行评估。团队会权衡准确性、响应时间和成本，因为最佳选择会因任务而异。
3. Turn user feedback into a self-training loop
3. 将用户反馈转化为自我训练循环
Once deployed, Fyxer’s system continues improving through real user feedback. When someone edits a draft before sending it, the difference between the original and final email shows which output they preferred. 一旦部署，Fyxer 的系统会通过真实的用户反馈持续改进。当用户在发送前编辑草稿时，原始邮件与最终邮件之间的差异就显示了他们更倾向于哪种输出。
Fyxer converts those comparisons into training data using Direct Preference Optimization (DPO). Instead of manually labeling every example, the model learns from pairs of outputs: the original draft and the user-edited version. Fyxer 使用直接偏好优化（DPO）将这些对比转化为训练数据。模型不再需要手动标注每一个示例，而是从成对的输出中学习：即原始草稿和用户编辑后的版本。
Every drafting change then goes through an A/B test. Fyxer ships the new version only when it produces a statistically significant improvement. 每一次草稿修改都会经过 A/B 测试。只有当新版本产生统计学意义上的显著改进时，Fyxer 才会发布该版本。