AI 提示詞: 專案管理

沒有AI輔助就管理專案,就像沒有GPS導航一樣——您會到達目的地,但會在繞路上浪費時間。這些提示詞幫助您規劃衝刺、分解複雜的交付物,並在時程失控前發現風險。已在ChatGPT、Gemini和Claude上測試,讓您為每個PM挑戰選擇最佳模型。

最近測試日期 Feb 15, 2026 · 模型: GPT-4o, Gemini 2.0, Claude 3.5 Sonnet, Grok 2
您想做的事 提示詞 最適合
規劃兩週的衝刺 衝刺設計 Claude
將大任務分解為子任務 任務分割 ChatGPT
識別和降低專案風險 風險雷達 Claude
繪製和管理利害關係人 利害關係人繪製 Gemini
撰寫清晰的專案狀態報告 狀態更新器 ChatGPT
引導有效的回顧會議 回顧會議主持 Gemini

衝刺設計

規劃兩週的衝刺

I need to plan a sprint for my team.

Project: [project name and goal]
Team members: [list roles and availability]
Backlog items: [list 10-15 user stories or tasks with estimates]
Sprint length: [1 week / 2 weeks]
Previous sprint velocity: [story points completed if known]

Create:
1. A prioritized sprint backlog with story point assignments
2. Task breakdown for the top 5 items (subtasks with time estimates)
3. Dependencies between tasks and the critical path
4. Risk flags — items likely to spill over and why
5. Daily standup focus areas for each day of the sprint
6. Sprint goal statement in one sentence

最適合: CLAUDE

Claude擅長識別任務之間隱藏的依賴關係,並提供現實的速度估算。即使您要求塞更多進去,也不會把衝刺塞得過滿。

已測試 Feb 15, 2026

進階技巧

包含最近3個衝刺的實際速度,而非理想的速度。AI知道團隊實際上能交付多少而非理想上能交付多少時,能規劃出好得多的衝刺。

任務分割

將大任務分解為子任務

I have a large task that needs to be broken down:

Task: [describe the big deliverable]
Deadline: [when it's due]
Team members available: [who can work on this]
Technical constraints: [tools, platforms, dependencies]
Definition of done: [what 'complete' looks like]

Break this into:
1. A work breakdown structure with 3 levels of detail
2. Estimated hours for each subtask
3. Which subtasks can run in parallel vs. must be sequential
4. The minimum viable version that could ship in half the time
5. Acceptance criteria for each subtask
6. A Gantt-style timeline showing the critical path

最適合: CHATGPT

ChatGPT產生最詳細和可執行的工作分解結構。子任務的粒度最適合Jira或Asana等工具——不會太粗也不會太細。

已測試 Feb 15, 2026

進階技巧

每次都問「最小可行版本」。讓AI區分真正需要的和錦上添花的,就解決了專案範圍界定最困難的部分。

風險雷達

識別和降低專案風險

Analyze the risks for this project:

Project: [name and description]
Timeline: [start and end dates]
Team: [size and experience level]
Budget: [total budget]
External dependencies: [vendors, APIs, approvals needed]
Stakeholders: [who cares about this project]

Provide:
1. A risk register with 8-10 risks ranked by likelihood and impact
2. For each risk: trigger condition, impact description, and mitigation plan
3. The top 3 risks that could kill the project entirely
4. Early warning indicators for each critical risk
5. A contingency plan for the single most likely risk
6. A risk review cadence recommendation

最適合: CLAUDE

Claude能識別其他模型遺漏的非顯著風險,特別是人的風險和組織政治。也提供最現實的影響評估而非通用的嚴重度評分。

已測試 Feb 15, 2026

進階技巧

包含上一個類似專案出了什麼問題。AI能捕捉專案之間的模式,並標記您已經經歷過但可能不會主動提及的風險。

利害關係人繪製

繪製和管理利害關係人

Help me manage stakeholders for this project:

Project: [name and one-sentence summary]
Stakeholders: [list each person with their role, department, and interest in the project]
Project phase: [kickoff / in-progress / nearing launch]
Biggest political challenge: [what makes this project politically tricky]

Create:
1. A stakeholder map: power vs. interest grid with each person placed
2. Communication plan: what each stakeholder needs to know, how often, and in what format
3. A RACI matrix for key project decisions
4. Scripts for managing a stakeholder who wants to expand scope
5. An escalation framework: when and how to escalate issues
6. A 'pre-mortem' — what stakeholder-related failure looks like and how to prevent it

最適合: GEMINI

Gemini建立最結構化的RACI矩陣和溝通計劃,視覺上清晰。利害關係人地圖可以不經重新格式化就向領導層展示。

已測試 Feb 15, 2026

進階技巧

在提示詞中對政治動態坦誠。AI只有知道誰擁有非正式權力、誰會抗拒、誰支持專案,才能幫助您引導利害關係人的政治。

狀態更新器

撰寫清晰的專案狀態報告

Generate a project status report from these details:

Project: [name]
Reporting period: [this week / this month]
Milestones completed: [list what shipped]
Milestones in progress: [list what's being worked on]
Blockers: [anything stuck and why]
Budget status: [on track / over / under]
Team morale: [any concerns]

Write:
1. An executive summary (3 sentences: overall status, biggest win, biggest concern)
2. A red/yellow/green status for: scope, timeline, budget, quality, team
3. Key accomplishments with business impact
4. Risks and blockers with owner and resolution plan
5. Next week's priorities
6. One decision needed from leadership (with recommendation)

最適合: CHATGPT

ChatGPT撰寫最精煉的高階主管導向狀態報告。從影響而非活動開始,是領導層真正想讀的內容。

已測試 Feb 15, 2026

進階技巧

永遠包含一個需要領導層做出的決定。只提供資訊的狀態報告會被忽視。要求行動的報告會被閱讀。讓對方容易說「是」或「不是」。

回顧會議主持

引導有效的回顧會議

Help me run a retrospective for my team:

Project/sprint: [what just finished]
Team size: [number of people]
What went well: [list 3-5 wins]
What didn't go well: [list 3-5 problems]
Previous retro action items: [what was decided last time]
Team dynamic: [any tension or morale issues]

Design a retro that includes:
1. A creative format beyond 'start/stop/continue' (with facilitation instructions)
2. 5 targeted questions to surface the real issues
3. A prioritization method for deciding which problems to fix
4. Action items template: specific, owned, and time-bound
5. A follow-up plan to ensure action items actually happen
6. A 'team health' pulse check with 5 dimensions to rate

最適合: GEMINI

Gemini設計包含清晰引導步驟的最有創意的回顧格式。行動項目模板夠具體,問責不是事後想法而是內建的。

已測試 Feb 15, 2026

進階技巧

在進行新的回顧之前,分享上次回顧的行動項目狀態。如果上次的項目沒有完成,團隊需要正視這個模式。

模型比較

基於實際測試結果 — 非假設推測。 查看測試方法

G

Gemini

Best for structured project artifacts like RACI matrices, stakeholder maps, and retro formats. Produces clean, presentation-ready outputs that work in Google Workspace immediately.

結果來源: Gemini 2.0 Flash · 已測試 Feb 15, 2026
C

ChatGPT

Best for status reports and work breakdown structures. Writes executive-ready summaries and creates task breakdowns at the right granularity for PM tools like Jira and Asana.

結果來源: GPT-4o · 已測試 Feb 15, 2026
C

Claude

Best for risk analysis and sprint planning. Identifies hidden dependencies and non-obvious risks, especially people and political dynamics. Gives honest assessments instead of optimistic projections.

結果來源: Claude 3.5 Sonnet · 已測試 Feb 15, 2026
G

Grok

Cuts through project management bureaucracy to identify what actually matters for delivery. Strong at calling out scope creep and unnecessary process, but less diplomatic in stakeholder communication advice than Claude.

結果來源: Grok 2 · 已測試 Feb 15, 2026

在 NailedIt 中試試

將上方的提示詞貼到 NailedIt,並排比較各模型的回應。

進階技巧

1

Plan for 70% capacity, not 100%. Teams that plan sprints at full capacity always miss targets. Buffer 30% for meetings, context switching, and surprise requests. AI will try to fill every hour — tell it your real available time.

2

Write the 'definition of done' before starting any task. Scope creep happens when 'done' isn't defined upfront. Ask AI to write acceptance criteria for every major deliverable before work begins.

3

Track decisions, not just tasks. Most project delays come from unmade decisions, not incomplete work. Ask AI to create a decision log alongside your task board.