Summary
Ramp launched AI Token Spend Management, giving finance teams visibility into AI model spending across providers. The tool covers Anthropic, OpenAI, Gemini, and Cursor, with sorting by provider, model, user, or API key. Average savings of 12% of monthly AI spending identified.
Key Points
- Problem: AI spending is a fast-growing cost area that finance teams struggle to track
- Why it's hard: token-based billing changes quickly; usage tied to API keys, not budget owners
- Ramp's solution: single dashboard across providers; sort by provider, model, user, API key, team
- Scale: developed with 1,300+ businesses managing 100T+ tokens/month
- Weekly briefing: explains major changes, flags spikes, suggests cost cuts
- Average savings: 12% of monthly AI spending
- Example 1: $1,600 weekend spending spike from one user
- Example 2: $3,000+ on Opus when Sonnet would have sufficed
- Example 3: Fast Mode increased costs 6x in one week unnecessarily
- Example 4: "prompt caching" surfaced by Ramp saved $10K/month (AngelList)
- Controls: soft/hard spending limits by user or API key; anomaly alerts via Slack/email
- Broader trend: AI governance becoming critical as adoption scales
- Comparable tools: CloudZero AI financial control plane; Parallel Works AI governance
- Key insight: finance teams need to distinguish between growth-linked spending and waste