---
title: "Check your AI spend assumptions"
canonical_url: "https://tokenobserve.com/tools/ai-spend-estimator"
published: 2026-10-15
last_updated: 2026-10-15
---

# Check your AI spend assumptions

Record what each source tells you, estimate token costs with your own rates, and see where evidence still needs checking. Each currency and cost basis keeps its own subtotal.

Calculations stay in this tab. Inputs are not sent to Token Observe or saved in browser storage. Export creates a CSV only when you choose it. Use labels rather than confidential account details.

Choose one month and add each distinct source once. Provider reports and invoices can describe the same usage. Their subtotals remain separate, as do gateway estimates. Rates are your assumptions per million tokens, not a live price feed. Include cache, reasoning, tool or other charges in your source records; the simple input/output estimate does not model them.

Subtotals include entered values only. Unknown amounts stay unknown. A reviewed checklist is a record of your checks, not a measured percentage of the AI estate.

For the same account, currency and completed UTC period, enter the first and second source totals. The result is second minus first. It explains no cause by itself. Compare gateway to provider, then provider to invoice as two separate checks.

## Review the evidence coverage

- I aligned the account, completed UTC period and currency.
- I checked collection freshness and partial or failed sources.
- I identified unconnected providers, subscriptions and bypass traffic.
- I checked cache pricing, credits, tax and other exclusions separately.
- I kept source records and a named person for unexplained differences.

- Work through a reconciliation: https://tokenobserve.com/guides/reconcile-ai-provider-spend
- Download the reconciliation worksheet: https://tokenobserve.com/downloads/ai-spend-reconciliation.csv
- Connect your first billing source: https://tokenobserve.com/get-started