@RampLabs: AI spend is measured in tokens, model calls, and dollars. However, none of these fields describe the actual work being …

X AI KOLs Following Tools

Summary

Ramp developed a semantic layer to attribute AI agent spend to objectives and outcomes, moving from monitoring spend to understanding AI ROI.

AI spend is measured in tokens, model calls, and dollars. However, none of these fields describe the actual work being done. At Ramp, we built a semantic layer that attributes agent spend to objectives and outcomes. This allowed us to go from monitoring AI spend to understanding AI ROI. Here's how we built it
Original Article
View Cached Full Text

Cached at: 09/02/26, 11:57 AM

AI spend is measured in tokens, model calls, and dollars. However, none of these fields describe the actual work being done.

At Ramp, we built a semantic layer that attributes agent spend to objectives and outcomes. This allowed us to go from monitoring AI spend to understanding AI ROI.

Here’s how we built it

We segmented agent traces into individual units of work. 17% of runs contained multiple work items and accounted for 58% of model spend. At the same time, 30% of runs with multiple linked sessions represented only one work item.

We did not start with a fixed taxonomy. Instead, we allowed our labels to emerge naturally from production agent traces. Because the taxonomy is learned, it can evolve with our usage.

The result is a record of each work item’s objective, outcome, and cost. This enables us to measure ROI for the agents we deploy and empowers intelligent resource allocation.

For full engineering writeup:

Similar Articles