A snapshot is your own OpenAI or Anthropic CSV, run through the same routing engine we use in production. If the saving is small, we say so.
Token counts and model names are enough. Prompt text is not required and we do not want it.
In the OpenAI platform, open Usage, export the last 30 days as CSV. Keep the model and token columns. That is the file we run.
From the Anthropic console usage view, export the same window. Claude model names in the file are enough for us to pick the right price ladder.
Rows analysed, current spend, spend if those rows had been routed, and the monthly saving — calculated on this page from your file.
The scan runs against the live Inferenz engine. Nothing is modelled, and a $0 result is a real result.
Need help exporting? Email hello@inferenz.com.au.