# AI Job Descriptions Feature Spec

## Problem
Site supervisors hate typing job posts on phones. Current process takes 5-10 minutes with full keyboard input. Need 2-3 tap solution.

## Solution
AI generates complete job descriptions from minimal inputs: trade + location + date.

## User Flow
1. Tap trade (carpenter, electrician, plumber, etc.)
2. Tap location (site address from saved list)
3. Tap date (calendar picker)
4. AI generates full description in 2 seconds
5. One-tap post (or quick edit option)

## Required Inputs (3 taps)
- **Trade**: Dropdown of 15 common construction trades
- **Location**: Pre-saved site addresses (can add new)
- **Date**: Calendar date picker

## AI-Generated Fields
- Job title (with trade + level)
- Detailed description with responsibilities
- Required skills/certifications
- Safety requirements (White Card, PPE)
- Pay range (based on trade + market data)
- Duration estimate
- Equipment/tools needed
- Physical requirements
- Experience level needed

## Prompt Engineering
```
Generate a construction job posting for a [TRADE] in [LOCATION] on [DATE].

Requirements:
- Include White Card requirement
- Mention PPE requirements
- Add physical demands (lifting, standing, etc.)
- Include pay range for Sydney market
- Specify typical duration for this trade
- List essential tools/equipment
- Add safety-focused language
- Keep it concise, mobile-friendly format
- Use Australian construction terminology
```

## Model Choice
**GPT-4o-mini** - $0.15 per 1M input tokens
- Fast enough for mobile (1-2 second response)
- Cost effective at scale (1000 posts = $0.15)
- Good quality for structured content
- No setup/maintenance vs local models

## Edge Cases
1. **New trade not in dropdown** → Add "Other" with manual input
2. **Multi-day jobs** → Auto-detect from description, offer date range
3. **Urgent/same-day jobs** → Flag as urgent, adjust wording
4. **Special requirements** → Quick add tags (heights, confined space, etc.)
5. **Repeat jobs** → One-tap repost from history

## Implementation Estimate
**2-3 weeks** for full implementation:
- Week 1: UI for 3-tap flow, location management
- Week 2: AI integration, prompt testing, field mapping
- Week 3: Edge cases, testing, deployment

## Success Metrics
- Time to post: <30 seconds (vs 5-10 minutes now)
- User satisfaction: 4+ star rating
- Post completion rate: >90%
- Reduction in support tickets about posting

## Future Enhancements
- Save AI-generated templates
- Voice input option
- Photo-based job description (snap site, AI describes)
- Smart scheduling based on trade availability