20%
Problem definition clarity.
問題界定清楚度。
Assignment
AI Agent Business Platform MVP Proposal
AI Agent 商業平台 MVP Proposal
The assignment extends the in-class project brief into a short proposal. The work is judged by clarity, feasibility, risk control, and business value.
課後作業是把課堂 project brief 擴充成簡短 proposal。評分重點是清楚、可執行、風險控管與商業價值。
Required Fields
A proposal that another team can continue.
讓另一個團隊可以接手的 proposal。
The proposal should make the idea discussable. A reader should understand the business problem, the first user, the data, the agent task, the review point, and the reason to fund the MVP.
Proposal 要讓構想變得可討論。讀者應該能看懂商業問題、第一個使用者、資料、Agent 任務、審核點，以及為什麼值得投入 MVP。
- Business problem and target users.
- Current workflow pain points.
- Proposed AI agent tasks.
- Data sources and data boundaries.
- Human-in-the-loop review points.
- MVP scope and 4-6 week roadmap.
- Risks and mitigation measures.
- Expected business value and success metrics.
- 商業問題與目標使用者。
- 現有流程痛點。
- AI Agent 任務設計。
- 資料來源與資料邊界。
- 人工審核點。
- MVP 範圍與 4-6 週導入規劃。
- 主要風險與對策。
- 預期商業價值與成效指標。
Evaluation
The rubric rewards management judgment.
評分重點是管理判斷。
Strong proposals explain why the selected process, data, review point, and MVP make sense.
好的 proposal 會說清楚為什麼選擇這個流程、這些資料、這個審核點與這個 MVP。
25%
Agent workflow design.
Agent 工作流程設計。
20%
Data boundaries and risk control.
資料邊界與風險控管。
20%
MVP and ROI logic.
MVP 與 ROI 邏輯。
15%
Communication and handoff quality.
表達與交接品質。
Check
Can a non-attendee understand the next step?
沒上課的人能否看懂下一步？
Weak Proposals to Avoid
避免這些薄弱寫法
- “Build an AI assistant” without naming the process.
- Listing tools without explaining data and review.
- Promising full automation before testing one small workflow.
- Using ROI language without a measurable baseline.
- 只說「做一個 AI 助理」，卻沒有說清楚流程。
- 只列工具，沒有交代資料與審核。
- 還沒測一條小流程，就承諾全面自動化。
- 講 ROI，卻沒有可比較的基準。