以前让 AI 写日报、做复盘、拆解笔记，每次都要重新解释一遍要求；工作里的计划和复盘又散在不同地方。这节课走一条完整的路：先把自己做事的方法写成 SOP，再把目标、项目、行动和复盘接成一个工作台。
琪琪琪
香港浸会大学教育硕士 AI教育行业产品经理 AI自媒体博主
Before, every time you asked AI to write a daily report, run a retrospective or break down a post, you had to re-explain the requirements; and your plans and reviews were scattered across different places. This course walks one complete path: write down how you actually work as an SOP, then connect goals, projects, actions and reviews into a workbench.
| 1 SOP card | 3 scenario prompts | 1 workbench blueprint | 3 deliverables |
|---|---|---|---|
| Six elements, fill in and use | Daily report / retrospective / post breakdown | The full loop from goal to feedback | A template pack + something you built + a workflow of your own |
Every session delivers three things: hands-on practice, notes and a workbench prototype, and homework.
The problem is not that AI is not smart enough — it is that you never stated the goal, the inputs, the decision rules or the completion criteria, and those live only in your head.
Write down the goal, inputs, steps, decision rules, output format and checking method for a repeated task so AI can run it again and again while a human can verify the result. The six-element formula — role + goal and boundaries + input spec + steps + decision constraints + output and acceptance — broken down element by element with a daily-report example, plus the four common problems and what each one means the SOP must fill in.
Daily report (from scattered notes to a fixed briefing), project retrospective (from experience to next-round actions), and social post breakdown (from observation to an original test) — the full prompts for all three, each with the specific error to hunt for and the rule to add.
Why a finished SOP still needs a workbench; the six-stage loop (OKR → Gantt → action list → today's three things → review → feedback into actions) with its key fields; the five-step build method; the copy-paste workbench prompt; and the cross-role reuse formula with a four-role comparison table.
This course has 3 lessons and 1 assignment, moving from understanding the concept to running three scenarios to building your own workbench.
Start from the problem worth solving; clarify the concept by comparing three levels (one-off question / AI SOP / workbench); break down the six-element formula (role, goal and boundaries, input spec, steps, decision constraints, output and acceptance) with a daily-report example for each; map four common problems to what the SOP must fill in; and land the most important line — a good SOP is never finished in one pass.
Full prompts and observation points for all three scenarios: A daily report, B project retrospective, C social post breakdown. Each gives demo material, steps, rules and output format, plus the v1 to v2 method of catching one specific error, adding one rule, and verifying with a second set of materials.
Why a finished SOP still needs a workbench (the SOP for one task is not the priority for one day); the six-stage loop and its key fields; the five-step build method; the copy-paste workbench prompt (flow and field tables first, generate the prototype only after confirmation); the cross-role reuse formula and a four-role comparison table.
In one sentence: what you take home today is not a universal prompt but a method — write your experience into an SOP, verify it with real work, then put repeated workflows into a workbench.
Plus three red lines from the lesson:
Get one thing running first, and the workbench will truly belong to you.
