
| Ramp AI Index businesses tracked | 85,000+ (author's claim) |
| AI spend growth in October | +1.65% - essentially flat (shown on screen) |
| US businesses on Ramp paying for AI | 56.8% (shown on screen) |
| Frontier models' share of total AI spend | 69.5% (shown on screen) |
| Customer Success share of AI spend | Grew from 0.8% to 13.3% between June and October (shown on screen) |
| Illustrative token price drop | $5.00 per million to $1.00 per million (author's claim) |
Who will this analysis save an hour?We’ll send the link with the “actually” takeaway — they won’t have to watch the whole video.
This happens because of 'trading down.' As cheaper models become more capable, businesses move their high-volume, routine tasks to these lower-cost options. You are likely getting more work done for the same price by utilizing more efficient unit economics.
Evals are standardized tests of the actual work you need done, judged against an acceptable business standard. Instead of trusting general benchmarks, you test models on your specific data to find the cheapest one that passes your requirements.
The most effective way is to implement a routing system. Use a cheap model for 90% of routine queries and only 'escalate' complex reasoning tasks to expensive frontier models. This prevents overpaying for simple tasks that don't require high-level intelligence.
Not necessarily, but the market is becoming tougher. While usage is exploding, model makers face more demands for discounts and competition from cheaper alternatives. The financial benefit is increasingly staying with the buyer rather than the supplier.
★ 9/10
★ 8
★ 8