Pre-conditions are free filtering
Documents skipped by pre-conditions cost zero credits. Documents skipped by the classifier cost 1–3 credits each. Filter early, filter cheap.
2 min read
Every stage in a bizSupply pipeline costs more than the one before it. The cheapest place to drop a document is the very beginning, before any AI call runs. That is what pre-conditions are for, and most teams underuse them.
Where credits start being spent
- Pre-conditions (file size, MIME type, content length, header pattern):
0 credits. - Classification (an AI call to label the document): 1–3 credits.
- Extraction (a larger AI call to pull structured fields): 3–25 credits.
- Aggregation (post-processing across documents): variable.
Use pre-conditions to encode common-sense rules
"PDFs only", "between 50KB and 5MB", "subject line contains invoice OR contract OR PO", "MIME type is application/pdf". These rules cost nothing to evaluate and can drop 30–60% of unrelated documents in a typical mailbox source.
Pattern: deny first, allow specific
Default to dropping documents that do not match a known shape, rather than letting everything through and trusting the classifier. The classifier is good but not free; pre-conditions are both fast and free, so they should carry as much of the filtering work as possible.
Still need help?
If this article does not solve it, the bizSupply team is one ticket away.