Build an AI cost-per-action monitoring dashboard

Description

Your AI feature works but costs are growing faster than users. This builds a cost-per-action dashboard — cost per user, cost per successful outcome, cost drift over time — so you catch runaway costs before the finance team does.

You are building a cost monitoring dashboard for {{ai_feature}}. Current monthly cost: {{current_cost}}.

## Metrics to track
### 1. Direct cost
- Total API calls per day
- Total tokens (input + output)
- Cost per call (model-weighted)
- Cost per user (monthly)

### 2. Cost per outcome
- Successful outcome rate
- Cost per successful outcome = direct cost ÷ successful outcomes
- Cost per attempted-but-failed outcome

### 3. Cost drift
- Week-over-week change
- Cost per user trend (is it rising?)
- Distribution analysis: top 10% of users consume what % of cost

### 4. Efficiency metrics
- Cache hit rate (if caching is used)
- Prompt size trend (prompts get longer over time)
- Model mix (if multiple models)

## Alerts
- Daily cost >X → review
- Cost per user >Y → investigate usage pattern
- Cost per successful outcome >Z → eval quality vs. cost tradeoff
- 7-day moving average up >20% week over week → explain or optimize

## Output
1. Dashboard spec
2. Alert thresholds for our scale
3. The one cost driver we'd expect to dominate
4. The optimization with highest expected savings

Customize This Prompt