TutorialsProject Retrospectives with AI
One GRAI four-step method, 3 hands-on prompts and 7 templates — turn a pile of scattered project records into a report and a slide outline you can present, in 30 minutes.
A retrospective is not about finding someone to blame. It is about taking the exam paper you wrote at kickoff, marking it again, and then asking one question: which rule needs to change.
This course walks one complete path: from records scattered across meeting minutes, group chats, task lists and data sheets, to a presentable retrospective report and a one-page slide outline in 30 minutes.
What you take away today
| 1 method | 3 hands-on runs | 7 templates | 1 retrospective canvas |
|---|---|---|---|
| GRAI: Goal → Result → Analysis → Insight | Materials in → Analysis out → Action forward | Sorted along the retrospective timeline, fill-in-the-blank ready | One table that runs the whole GRAI cycle |
The three-part promise: live practice + template files (prompt pack) + homework.
You may have run into this
- The activity ended a while ago and the retrospective keeps slipping — minutes, chats and task sheets are spread across five places, and rebuilding the timeline by hand takes two hours;
- The retrospective meeting drifts into a contest of impressions: "I think we underspent on promotion" versus "I think the content was weak", with no way to settle it;
- Analysis stops at "be more careful next time" — it sounds right, but the same pit is hit again next round;
- You finally finish the report, and your manager asks "so what is the conclusion and what changes next quarter?" while the report is all process description.
The problem is not that you cannot do a retrospective. It is that the process records were never rebuilt into a timeline, and goals and results were never laid out on the same table — without those two, the rest of the discussion can only run on feeling.
What you will learn
Start here: what a retrospective really does
Four concepts explained in plain language — what a retrospective is, the GRAI four-step method, workflows versus one-off conversations, and materials versus information gaps; then seven high-frequency scenarios, one by one. It ends with the course formula (role + task + context + format + constraints) and the single most valuable follow-up trick.
Hands-on: materials in → analysis out → action forward
Using the 'AI Open Class · Back-to-School Special' case, you run three prompts live — rebuild five scattered materials into a timeline and flag contradictions and gaps, restore the goals and compare against actuals, then run a three-layer root-cause analysis and turn it into an action plan. Each stage comes with the full prompt and what to watch for.
Take-home practice: seven templates and the red lines
Seven prompt templates arranged along the retrospective timeline (prep → G → R → A → I → reporting, plus a one-line starter), fill-in-the-blank ready; a blank retrospective canvas that closes out any project; a tool selection table; and finally the four red lines, especially the one about blaming people instead of processes.
Who this is for
Course structure
Three lessons plus one assignment: from "what a retrospective really reviews" to running a complete retrospective on five types of materials live, and finally fixing the process into your own retrospective workflow.
Start here: what a retrospective really does
What a retrospective is (people-free), the GRAI four-step method, workflows versus one-off conversations, the five material types and information gaps; seven high-frequency scenarios; and the course formula (role + task + context + format + constraints) with the most valuable follow-up trick.
Hands-on: materials in → analysis out → action forward
Using the "AI Open Class · Back-to-School Special" case, three prompts run in relay: prompt A rebuilds five scattered materials into a timeline with contradictions and gaps flagged; prompt B restores the goals and their measurement basis and outputs a goal-versus-actual table; prompt C runs the three-layer root-cause analysis and produces experience assets, an action plan and a presentable report.
Take-home practice: seven templates and the red lines
Seven copy-ready prompt templates arranged along the retrospective timeline, each marked with its purpose and when to use it; a blank project retrospective canvas; a tool selection table; and the four red lines, especially this course's own — root causes point at processes and mechanisms, never at people.
Course materials
This course ships with the five source materials of the main case (AI Open Class · Back-to-School Special; fictional demo data, all five material types complete):
- Two meeting minutes: the kickoff meeting (goals, measurement basis, roles, schedule) and the live-day stand-up (post-live huddle and open issues);
- Group chat excerpt: 20 messages tied to key milestones between 9/2 and 9/21;
- Task list: 12 tasks with owners, planned and actual completion dates, status and notes;
- Activity data: funnel overview, channel breakdown, live-show metrics, budget, and a time-loss record.
Download (used in the live practice and in part one of the assignment):
- AI项目复盘-01会议纪要-项目启动会.md (meeting minutes)
- AI项目复盘-02会议纪要-直播当天站会.md (meeting minutes)
- AI项目复盘-03项目群聊记录节选.md (group chat excerpt)
- AI项目复盘-04任务清单.csv (task list)
- AI项目复盘-05活动数据.md (activity data)
The principle behind it all
In one sentence: AI writes the first draft and finds what you missed; you judge and decide. Treat it as a very smart subordinate — you assign the work, and you make the call.
Four red lines to go with it:
- AI makes things up with a straight face: numbers, dates and names must be checked by hand — every figure in a retrospective has to trace back to the source data sheet;
- Never feed it confidential or sensitive information: anonymise first, or do not feed it at all — client lists, internal pricing and unpublished data all get anonymised;
- The output is not the final version: your name is on it, and so is the responsibility — review it once yourself before it goes out;
- Retrospectives target processes, not people: root causes must point at processes and mechanisms, and AI output must never become a case file against a colleague; before it reaches your manager, confirm the evidence chain behind every root cause.
Assignment: Compress a Meeting into One Page
Three levels in one assignment — reproduce the three runs with the sample transcript, run a real meeting and time yourself, then fix your own minutes workflow. All three required.
Start Here: What a Retrospective Really Does
Four core concepts (retrospectives, the GRAI method, workflows versus one-off conversations, materials and information gaps), seven scenarios, and the course formula.