Summarize 1.2M care calls a month
The batch discount is the go/no go on the whole program.
Open the live lab · preloaded to this scenario
Token Economics Simulator
Context
A telecom summarizes ~1.2M care center calls a month for QA and coaching, an offline, non urgent workload. Because it's async, most of it can run through the batch API at a steep discount.
The decision
Batching is the go/no go: at this volume the batch discount is what moves the program from 'too expensive' to fundable, the workload is async, so there's no reason to pay real time rates.
What most miss
People price offline workloads at real time rates. Summarization is async; routing it through batch is money left on the table if you don't.
Stakes
Priced at real time rates, a 1.2M call summarization program doesn't clear its business case.
Firsthand · Agent Architecture and Protocol Strategy Artifacts · verified 2026-07-03
Sources: Telecom care-center call summarization at scale, firsthand (Verizon); Batch-inference discount economics