D10Intermediate
Token Cost Optimization
30 minEvery AI feature
Format: Learn how to reduce token consumption in AI API calls (save money).
Optimization strategies:
| Strategy | Savings | Example |
|---|---|---|
| Shorten system prompt | 10-30% | Remove unnecessary instructions |
| Cache common answers | 50-80% | Don't call API for the same questions |
| Tiered models | 60-70% | Use Haiku for simple questions, Sonnet for complex ones |
| Limit output length | 20-40% | "Answer in 50 words or less" |
| Batch processing | 30-50% | Process multiple questions in one request |
Exercise: Estimate the monthly AI cost for the following scenario:
AI customer service bot:
- 500 conversations per day
- Average 6 rounds per conversation
- Per round: 300 input tokens + 200 output tokens
- Using Claude Sonnet
Your estimate: $___/month
Then: Propose 3 ways to reduce costs by 50%