AI workflow acceptance and exception log
Keep samples, versions, human baselines, error types and review effort together for a useful trial review.
Original guidance and working materials. Validate against your business; these are not measured project results or delivery commitments.
Test conditions
- Business task and input scope: ____
- Product or model, prompt or rule versions: ____
- Document, template and data versions: ____
- Test date, operator and business reviewer: ____
- Sample sources and permitted use: ____
Record each sample
| Sample ID | Type | Expected result and evidence | Actual result | Error or omission | Human action | Time | Review outcome |
|---|---|---|---|---|---|---|---|
| Complete | Normal or exception | Complete | Complete | Complete | Complete | Complete | Complete |
Preserve the denominator
Separate normal and exceptional samples. Include incomplete runs, failures and human intervention. Record preparation, checking and rework for both workflows. A fast demonstration is not a general result.
Synthetic text samples
The additional JSON file contains fictional order text with complete fields, a missing date, a duplicate and a unit difference. It tests field and exception rules. It contains no PDF files and cannot establish PDF recognition performance.
Record the decision
- Scope that can continue: ____
- Changes, missing data and retests needed: ____
- Owner and next review date: ____
- Evidence for acceptance or stopping: ____
Download working materials
Editable templates and synthetic examples. No real customer information.
Editable Markdown templateDownload ↓Synthetic order text samples (JSON)Download ↓