Use-case preset
Sales call analysis cost calculator
Sales call transcript to talk-track summary, objections, and next steps.
A 30–60 minute sales call transcript is typically 8–20k tokens. The job is to extract a structured summary: talk-track breakdown, objections raised, commitments made, and recommended next steps. The 90/10 ratio is accurate — the transcript dominates, the structured output is compact. Context is 32k to handle longer enterprise calls or multi-participant calls.
Batch processing fits naturally; reps review AI summaries after the call, not during. Cached prompt percent is near zero since every transcript is unique. The real ROI question is extraction accuracy on objections and action items — these are the high-value fields. Run precision/recall benchmarks against human-labeled samples before automating the CRM write-back. Token cost is low: at $1/M input and 10k average transcript length, 1,000 calls/month runs about $9 in input costs.