Human in the Loop
Route uncertain decisions to people, automatically
When your model is confident, accept its answer. When it is not, call Cashilly and get a human judgment back through the same API. Your pipeline never stops to wait for a person to be found.
What workers do
- Verify a low-confidence classification
- Resolve an ambiguous extraction
- Confirm a decision before it reaches a customer
- Review an edge case your rules did not cover
What you upload
- any fields your model was reasoning over
Any format: CSV, TSV, JSON, JSONL, XLSX or TXT. Columns are detected automatically.
Common questions
- How do I integrate this?
- POST the uncertain item to /api/v1/judgments and either poll the judgment id or receive a webhook when it resolves.
- How fast is it?
- Turnaround depends on how many qualified workers are available for your requirements at that moment. Every project shows a live estimate, and the API returns one at submission.