过去，获取信息本身就是一种成本。查资料、做分析、请教他人，往往意味着时间与经验的积累。因此，一个人的能力，在很大程度上体现为他能掌握多少信息、沉淀多少知识，并把这些信息转化为判断。
In the past, access to information came with a real cost. Looking up materials, doing analysis, and consulting others all required time and accumulated experience. As a result, a person's capability was often reflected in how much information they could master, how much knowledge they could retain, and how effectively they could turn that information into judgment.
而今天，随着AI的发展，获取初步答案变得前所未有地高效。很多过去需要反复推敲的工作，如今只需一次提问，就可以得到一个结构完整、看似成熟的结果。
Today, with the development of AI, obtaining an initial answer has become more efficient than ever. Many tasks that once required repeated thinking and refinement can now produce a structured and seemingly mature result from a single prompt.
我们曾构建过一个用于模具测试数据诊断（MTA）的AI应用。在这个场景里，工程师不再需要逐条人工分析测试数据，而是可以让AI基于历史数据给出诊断建议。一旦输入某一条测试结果，系统会自动提取相同模具、相同零件的历史记录，整理并组合相关数据，再通过结构化提示词发送给AI大模型进行分析，最终输出标准化的问题诊断结果。这个过程看似是在“获取答案”，但真正影响结果质量的，不只是AI本身的能力，更是我们如何把业务问题、数据背景和判断标准组织成AI能够理解的结构。
We once built an AI application for mold testing analysis (MTA) data diagnostics. In this scenario, engineers no longer needed to analyze each test record manually; instead, AI could provide diagnostic suggestions based on historical data. Once a specific test result was entered, the system automatically retrieved historical records for the same mold and the same part, organized and combined the relevant data, and then sent it to a large AI model through structured prompts for analysis. The final output was a standardized diagnostic result. This process may look like simply “getting an answer,” but the quality of the result depends not only on the AI model itself, but also on how we organize the business question, data context, and judgment criteria into a structure the AI can understand.
当答案越来越容易获得，问题本身的定义反而变得更加重要。
As answers become easier to obtain, the way we define the question becomes even more important.
首先，思考问题的方式正在从“追求答案”转向“明确问题”。当AI可以同时给出多个看似合理的结果时，真正决定答案质量的，不只是模型能力，更是提问者是否把问题说清楚。很多时候，不是我们没有答案，而是我们还没有把问题想清楚。如果只是简单地把一行测试数据交给AI，并问：“这条数据有没有问题？”AI往往只能给出较为泛化，甚至不稳定的判断。但在实际应用中，我们会先把问题拆解并结构化：让AI关注多个模穴之间的差异、历史数据变化的趋势、材料或机器测试条件的变化，以及这些因素之间可能存在的关联。我们甚至会人为定义一些诊断规则，作为AI分析时的参考。当这些观察维度和诊断规则被清晰定义后，再交由AI处理，输出结果的稳定性和可解释性都会显著提升。
First, the way we think is shifting from “pursuing answers” to “clarifying questions.” When AI can generate several seemingly reasonable results at the same time, the quality of the answer is determined not only by the model, but also by whether the person asking the question has made the problem clear. In many cases, the issue is not that we lack answers; it is that we have not yet clarified the question. If we simply give AI one row of test data and ask, “Is there a problem with this data?”, AI will often provide a generic or even unstable judgment. In real applications, we first break down and structure the question: we guide AI to focus on differences across mold cavities, trends in historical data, changes in material or machine testing conditions, and the possible relationships among these factors. We may even define diagnostic rules manually to serve as references for AI analysis. Once these observation dimensions and diagnostic rules are clearly defined, AI can produce results that are much more stable and explainable.
模糊的问题，只会得到表面的回答；结构清晰、目标明确的问题，才能获得有价值的结果。
A vague question can only lead to a surface-level answer. A well-structured question with a clear objective is what makes a result truly valuable.
其次，人的角色正在从执行者转向设计者，工作方式也在从“自己完成”转向“设计让事情被完成”。过去，我们强调亲力亲为，从数据收集到结果输出，很多环节都依赖人工完成。而在MTA&AI应用中，历史数据的自动提取与组合、AI分析诊断的调用、诊断结果的标准化回写，都可以被设计成流程的一部分。信息整理、内容生成、基础分析等环节，正在越来越多地由AI承担。
Second, the human role is shifting from executor to designer, and the way work gets done is moving from “doing it ourselves” to “designing how it gets done.” In the past, we emphasized hands-on execution, and many steps from data collection to result delivery relied on manual work. In the MTA + AI application, however, the automatic retrieval and combination of historical data, the invocation of AI analysis, and the standardized writing-back of diagnostic results can all be designed into the workflow. Tasks such as information organization, content generation, and basic analysis are increasingly being handled by AI.
真正重要的能力，不再只是把事情做完，而是能否把事情设计成可以被高效、稳定、可复用地完成。一个人的工作价值，不只体现在做了多少，更体现在能否让同样的事情以更高质量、更低成本持续完成。
The truly important capability is no longer just getting things done, but designing work so it can be completed efficiently, consistently, and repeatedly. A person's value at work is not only reflected in how much they do, but also in whether they can enable the same work to be done with higher quality and lower cost over time.
与此同时，决策方式也在发生转变：从单一经验，走向人机结合。在MTA应用中，AI可以基于历史数据快速给出诊断结论，例如可能的异常环节或风险点。但在实际业务中，这些结果并不会被直接采纳，而是作为决策参考。工程师仍然需要结合生产背景、设备状态、工艺知识和现场经验，对结果进行判断与确认。AI负责提供多个“可能性”，人负责选择“最合理的解释”，并决定下一步行动。因此，判断不会被替代，但判断的方式正在升级。更有效的决策，不再只来自单一经验，而来自多种可能之间的选择、验证与取舍。AI提供广度，人类提供方向。
At the same time, decision-making is also changing: from relying on individual experience to combining human judgment with AI support. In the MTA application, AI can quickly generate diagnostic conclusions based on historical data, such as possible abnormal steps or risk points. But in real business contexts, these results are not adopted directly; they serve as decision references. Engineers still need to evaluate and confirm the results based on production context, equipment status, process knowledge, and on-site experience. AI provides multiple possibilities, while humans select the most reasonable explanation and decide the next action. Judgment will not be replaced, but the way we make judgments is being upgraded. More effective decisions no longer come only from individual experience; they come from selecting, validating, and weighing multiple possibilities. AI provides breadth, and humans provide direction.
在这样的变化下，能力结构也在悄然重组。一些曾经非常重要的能力，例如信息整理、基础写作、简单分析，正在逐渐变成“默认能力”；而理解问题、结构化思考、流程设计和判断决策，则变得越来越关键。
With these changes, the structure of capability is also being reshaped. Some abilities that used to be highly important, such as organizing information, basic writing, and simple analysis, are gradually becoming default capabilities. Meanwhile, understanding problems, thinking structurally, designing workflows, and making sound judgments are becoming increasingly critical.
AI降低了执行的门槛，但也抬高了思考的门槛。
AI lowers the threshold for execution, but it also raises the threshold for thinking.
面对这样的转变，关键不在于是否使用AI，而在于我们是否真正改变了思考方式。当获取答案变得容易，人类的价值将不再主要体现在“给出答案”，而在于“提出什么问题，以及如何做出选择”。
In the face of this shift, the key question is not whether we use AI, but whether we truly change the way we think. When answers become easy to obtain, human value will no longer lie mainly in giving answers, but in asking the right questions and making the right choices.
未来的差距，不只是效率的差距，更是思考方式的差距。
The gap in the future will not only be a gap in efficiency, but also a gap in ways of thinking.
谁能更早从“寻找答案”转向“定义问题”，从“亲自完成”转向“设计完成”，谁就更容易在AI时代找到自己的位置。
Those who can shift earlier from “searching for answers” to “defining problems,” and from “doing the work themselves” to “designing how work gets done,” will be better positioned to find their place in the age of AI.