Author: Zhu Anfeng

When planning a product from 0 to 1—especially in government/enterprise domains—product managers often face a recurring problem: business data fields are unclear. In early iterations, designing a prototype frequently runs into this:

• the business is not mature, so nobody can define the data structure precisely
• engineering and data teams have not joined yet, so detailed data support is missing
• the UI needs realistic data, otherwise it looks empty

At this stage, PMs have to research and infer a reasonable set of fields. Otherwise the prototype becomes a shell: it fails to communicate requirements and may even impact decision-making.

💡 How can AI help?

AI can act as an intelligent data assistant and help PMs:

✅ infer likely fields based on comparable products and domain logic
✅ quickly generate sample records so prototypes look real and discussions become easier
✅ reference best practices to avoid missing critical fields

🎯 An example:

For a real requirement, I planned a module for "Smart City: Crowd Gathering Emergency Management" with:

1. pre-event, in-event, and post-event handling
2. for each function: feature description, interaction flow, data fields, inputs and outputs
3. for data fields: 10+ fields and 3 sample rows

👉 Below is what an AI produced. Different models (DeepSeek, ChatGPT, etc.) give similar answers, and they are at least a helpful starting point.

---

Product Module: Smart City Crowd Gathering Emergency Management System

1. Pre-Event Warning (Risk Prediction and Prevention)

Feature description

• Combine historical data, real-time monitoring, and AI prediction models to assess crowd-gathering risk and warn early.
• Define crowd thresholds for different scenarios (concerts, malls, metro stations) and notify relevant departments in advance.

Interaction

• Admin users view traffic trends of hotspots and configure warning rules.
• Alerts are pushed to emergency management departments, merchants, and relevant managers via an app.

Data fields

1. Location ID (unique identifier)
2. Location name (mall, station, etc.)
3. Real-time crowd count
4. Crowd growth rate (per minute/hour)
5. Max capacity (safe upper bound)
6. Weather
7. Historical crowd data (similar time windows)
8. Incident history
9. Safety level (low/medium/high)
10. Warning status (normal/notice/warning)

Sample data

Location ID	Location name	Real-time crowd count	Growth rate	Max capacity	Weather	Historical crowd data	Incident history	Safety level	Warning status
A001	XX Mall	1200	+5%	5000	Sunny	1100	None	Low	Normal
A002	XX Metro Station	4500	+8%	6000	Cloudy	4000	Congestion	Medium	Notice
A003	XX Concert Venue	9800	+12%	10000	Rain	9200	Stampede	High	Warning

---

2. In-Event Emergency Response

Feature description

• When anomalies occur, automatically trigger emergency response and dispatch resources.
• Provide best evacuation routes and coordinate with broadcast systems and command centers.

Interaction

• Show incident location, affected area, crowd density, etc.
• Provide AI-recommended actions such as evacuation routes and additional security staffing.

Data fields

1. Incident ID
2. Incident type (stampede, fire, violence, etc.)
3. Time (to the second)
4. Location (linked to Location ID)
5. Affected crowd size (estimated)
6. Response level (low/medium/high)
7. Recommended actions
8. Dispatched resources (responders, equipment, etc.)
9. Live video feed
10. Status (in progress/resolved)

Sample data

Incident ID	Type	Time	Location	Affected crowd	Response level	Recommended actions	Resources	Live feed	Status
E001	Stampede	18:30:12	A003	200	High	Evacuate via south gate	10 security staff	Live	In progress
E002	Fire	19:15:50	A002	500	High	Dispatch firefighters	3 fire trucks	Live	In progress
E003	Violence	20:05:30	A001	50	Medium	Control individuals	5 police officers	Live	Resolved

---

3. Post-Event Review and Optimization

Feature description

• Record the full incident lifecycle, evaluate response effectiveness, and optimize emergency plans.
• Generate statistical reports for decision-makers.

Interaction

• Incident replay with key decision timestamps.
• Analytics modules to evaluate response efficiency.

Data fields

1. Incident ID (linked)
2. Incident type (linked)
3. Response time (from detection to handling)
4. Actions taken (final)
5. Effectiveness (effective/partly effective/ineffective)
6. Casualties
7. Economic loss (estimate)
8. Optimization suggestions (auto-analysis)
9. Responsible unit
10. Review status (not reviewed/reviewed)

Sample data

Incident ID	Type	Response time	Actions taken	Effectiveness	Casualties	Economic loss	Suggestions	Responsible unit	Review status
E001	Stampede	5 min	Evacuate via south gate	Effective	0	0	Add more corridors	Security company	Reviewed
E002	Fire	8 min	Dispatch firefighters	Partly effective	2	100k	Inspect wiring earlier	Fire brigade	Reviewed
E003	Violence	3 min	Control individuals	Effective	0	0	Increase patrols	Police	Reviewed

---

Summary

This smart-city emergency system covers pre-warning, in-event response, and post-event review. With data analysis, AI prediction, and real-time monitoring, it improves safety in crowd-gathering areas and makes emergency management more effective.

With a field list like this, product teammates can quickly build a fuller-looking prototype. It may be a rough start, but someone has to do 0-to-1 work, right?

Previous pageThree Skills for the Future Workplace

Next pageHiking Jingde Xishan and a Study of a 1506 Stone Pavilion Inscription