Agent spend forecasting: the ledger you're not keeping
Forecasting agent spend isn't predicting the future — it's reading the audit log. The three inputs that make the forecast real.
Most AI spend forecasts are vibes: take last month, add growth, round up. The audit log makes it arithmetic instead.
The three inputs
Per-agent actuals — what each agent really spends, from the log. Volume trend — how calls grow per agent and per category. Ceiling proximity — how close agents run to their caps, which tells you whether the forecast is a floor or a ceiling.
The forecast
Next month ≈ per-agent actuals × volume trend, bounded by ceilings. That's not prediction — it's the ledger doing its job. Track it → The monthly cost report turns the log into the forecast.