工作中的 AI 基础
每个人都从这条核心课程开始——每次 AI 推行都会遇到的五个问题,按顺序逐一修复。
55 节课 · 222 分钟
预计阅读时间。
W · M0–M2
Get started and choose the work
Understand AI, select a suitable work step, and prepare approved information.
Deliverable: AI Trial Card, reviewed draft, and source register.
- 5 节课 · 20 分钟
W为什么这很重要
你即将学到的大多数内容都是技能——但技能只有在你真正想使用它时才会真正掌握。这个模块在讲“怎么做”之前先讲“为什么”:关于为何大多数公司在 AI 落地上做错了的真实数据、两个真实案例(一个成功,一个失败),以及这个项目对你的真正要求。
- 3 节课 · 14 分钟
M0先找准方向
还没认真用过 AI 的人,要么把它当成万能的查询工具来信任(第一次它编造内容时就会被坑),要么把它当玩具一样轻视(白白浪费掉本课程剩下的部分)。这个模块在教任何技巧之前,先纠正一个预期:它是一个非常出色的语言预测器,不是一个无所不知的数据库。
- 4 节课 · 11 分钟
M1定框 — 选择要做的工作
初学者最常见的两个错误:把不可逆或高风险的决定交给 AI 而吃了亏,或者相反——因为“AI 大概做不了这个”而事事亲力亲为,结果什么也没省下来。这个模块提供一套打分方法(频率 × 耗时 × 风险),而不是靠猜。
- 5 节课 · 20 分钟
M2接地 — 给它真实的上下文
有人试用几次 AI 后就放弃,最常见的原因是:它对自己公司的事情回答得很流畅,却是错的——而他们不知道该怎么解决。其他课程都把这归入“把提示词写好一点”;在这里它是独立的一个模块,因为限制不在于措辞,而在于模型从未见过你公司的文档。
M3–M5
Build a reusable Skill
Write instructions, evaluate representative cases, and package a tested approach.
Deliverable: Skill specification, evaluation set, version, and before/after evidence.
- 5 节课 · 16 分钟
M3定规 — 写一个可复用的请求
如果每次使用 AI 都是临时发挥,生产力其实并没有真正提升——时间只是从打字变成了猜测。这个模块把一个重复出现的任务变成一个公式:写一次,多次复用,每次结果的形状都一样。
- 4 节课 · 14 分钟
M4核验 — 在合适的深度上核查
真正的风险不是 AI 会出错——而是你不知道它什么时候出错,并且核查错了对象(对低风险工作反复审视,却跳过真正重要的部分)。这个模块用一套方法取代直觉,来决定某项任务值得多少核查。
- 5 节课 · 19 分钟
M5创建并测试 Skill
为一项任务定义可复用的 Skill，用常见情况和例外情况进行测试，并记录版本。在决定是否继续试用、分享或组合进工作流前，比较质量、全流程耗时、返工和成本。
M6–M8
Connect workflow, knowledge, and tools
Compose Skills, govern Knowledge/RAG sources, and connect systems within a defined scope.
Deliverable: Workflow Blueprint, Knowledge Charter, and Tool Charter.
M9–M11
Deploy and operate an agent
Choose an architecture, run a pilot, and decide whether to scale, redesign, or stop.
Deliverable: Architecture decision, Agent Pilot Charter, and operating review.
- 3 节课 · 15 分钟
M9Decide whether an agent is needed
Choose the simplest approach that meets the business need before any agent pilot.
- 3 节课 · 13 分钟
M10Run an agent pilot responsibly
Design a bounded pilot before any wider release.
- 2 节课 · 13 分钟
M11Scale, redesign, or retire with evidence
Use pilot evidence to make the next operating decision.
M12
Enterprise Training, Consulting, and Implementation
MasterOS goes beyond training people to use AI. We work with businesses to assess current operations, identify gaps, design solutions, and implement them. M12 connects the self-paced program to services tailored to your organization's goals, processes, and systems.
- Assess processes, data, and team capabilities; prioritize AI opportunities and plan around business goals.
- Provide role-based training, develop AI Champions, and coach teams on their own work.
- Design and implement Skills, workflows, Knowledge/RAG, tool integrations, and AI agents for the agreed needs.
- Establish governance: access, approval responsibilities, quality checks, and incident procedures.
- Measure quality, cost, and operational results; hand over, maintain, and expand across departments.
Tailored enterprise services; excluded from self-paced lesson counts and reading time. Scope, deliverables, and responsibilities are agreed before implementation.
Discuss your business needsApply the program to your role
After the main learning path, choose a specialization for your department's work or your responsibilities in an AI rollout.