Enrich a CSV with AI without losing the original data
Enrichment should add information to a record while keeping the original evidence intact. A useful output lets someone distinguish what the input actually said from what a model inferred.
Use new columns for derived values
For a contact list, a model might normalize job titles or classify a role as engineering, operations, or sales. Those are derived labels. A guessed employer, email address, or company size is a different kind of claim and should not be presented as verified data. Name fields so the distinction survives when the file leaves your workflow.
- 1Preserve source IDs and all original fields…Preserve source IDs and all original fields needed for review.
- 2Add explicit columns for normalized or inferred…Add explicit columns for normalized or inferred values.
- 3Reject unsupported factual additions instead of filling…Reject unsupported factual additions instead of filling every blank.
Try it on a small example
- Preserve source IDs and all original fields needed for review.
- Add explicit columns for normalized or inferred values.
- Reject unsupported factual additions instead of filling every blank.
What to verify
Check row counts before and after enrichment, plus the behavior of quoted commas and multiline cells in the exported file. A visually tidy preview does not prove the CSV round trip worked. Baleybots pipelines can read uploaded CSVs and produce sink CSVs; inspect both the original record and the processor result when validating an enrichment. Keep uncertain values nullable or route them to review.
Product details checked against the Public API reference on October 4, 2026. These guides describe documented behavior; availability depends on your account and the deployed service. Baleybots is in invite-only beta.