ZS Shanghai GenAI Hackathon

The three ingredients

三部分 · 权重相同 · 最后加总

⅓⚙️

Part 1 · Base Score

Impact × Difficulty

基础分 · 高价值 + 高难度

Every use case carries a base score from two things: how much business value it unlocks, and how technically hard it is. High-value, ambitious builds start ahead.

Formula · 计算公式

Base = Business × 0.7
+ Tech × 0.3

🏥 Business impact商业影响 (0–5)70%

⚙️ Technical difficulty技术难度 · 越难越高30%

🏥 Business 70%⚙️ Tech 30%

Result is normalized to a 0–100 base score.

⅓🧑‍⚖️

Part 2 · Judge Score

The expert verdict

评委打分 · 专业评审

Each judge rates the demo on 6 weighted criteria in the Judge Console. Scores are weighted, averaged across all judges, then normalized to 0–100.

6 criteria · equal weight (⅙) · 细则见下 👇

Relevance to consulting workflows咨询核心场景契合度⅙

Time saving / rework reduction提效与减少返工⅙

Depth of thinking & storytelling思考深度与叙事力⅙

Reusability across projects跨项目复用性⅙

Responsible use & QC责任使用与质控⅙

Innovation / client "wow"创新与客户惊喜度⅙

⅓🗳️

Part 3 · Team Vote

Peer love

团队投票 · 大家说了算

Everyone in the room is a voter. Spend your three votes on the teams that wowed you most — the crowd's favourites climb the board.

×3votes per person每人 3 票

✓✓✓

🎯 spend on your favourite teams 🔒 100% anonymous · 匿名 🏆 most votes climb the board

Judge rubric · 评分细则

6 criteria · equal weight (⅙ each) · click to expand 点击展开 ✅/❌

1Relevance to consulting workflows咨询核心场景契合度⅙+

是否切中咨询真实、高频的痛点？Does it target a real, frequent consulting task?

✅ Good — Targets a real, recurring task in ZS engagements; a consultant could use it next Monday. 切中真实高频场景，落地清晰。

❌ Not — Generic AI demo or toy problem with no clear tie to consulting work. 泛泛 demo / 玩具问题，与咨询无关。

2Time saving / rework reduction提效与减少返工⅙+

是否实质节省工时 / 减少返工？Does it meaningfully cut hours or rework?

✅ Good — Clear before/after time saving; automates the tedious bulk; cuts review cycles. 有量化提效，自动化最枯燥的部分。

❌ Not — Marginal or unclear savings; adds steps or needs heavy manual cleanup. 提效有限，甚至反增负担。

3Depth of thinking & storytelling思考深度与叙事力⅙+

推理质量 + 表达是否清晰有力？Quality of reasoning and how well it's told.

✅ Good — Thoughtful framing, handles edge cases, clear "what / so-what / caveat" narrative — insight, not raw output. 有洞见、叙事清晰、考虑边界。

❌ Not — Shallow prompt wrapper that just dumps model output; no structure or insight. 套壳甩输出，无结构无洞见。

4Reusability across projects跨项目复用性⅙+

能否被其他项目 / 团队复用？Can others reuse it beyond one case?

✅ Good — Generalizable & configurable across clients/data/teams; documented; templated. 可泛化、可配置、有文档。

❌ Not — One-off hack hard-coded to a single file/client; brittle; can't be handed off. 一次性、写死、难交接。

5Responsible use & quality control责任使用与质控⅙+

是否安全、准确、可治理？Safe, accurate, governed?

✅ Good — Handles privacy/PII, cites sources, flags uncertainty, keeps a human in the loop. 重隐私、标来源、提示不确定性、有人工校验。

❌ Not — Leaks sensitive data; no accuracy checks; treats AI output as truth; no guardrails. 外发敏感数据、无校验、无护栏。

6Innovation / client "wow" factor创新与客户惊喜度⅙+

原创性与打动客户的潜力？Originality and client delight.

✅ Good — Novel approach or combination; makes judges go "I hadn't thought of that"; clear client delight. 新颖、令人眼前一亮、有惊艳潜力。

❌ Not — Obvious or done-before; an incremental rebuild of an existing tool; no wow. 显而易见 / 重复造轮子，毫无惊喜。

Scoring Rules 🏆

One score, three equal slices

The three ingredients

Impact × Difficulty

The expert verdict

Peer love

Judge rubric · 评分细则

How it adds up