INI-003
·Aprovada p/ PilotoAutomatic invoice classification
Camila Reis · FinanceDocument processingCriticality HighAutonomy: Executes with review
84Overall
Automatic evaluation
Business Value88
AI Quality82
Security76
Privacy92
Opinion
Classic, mature use case. Clear ROI, low privacy risk, structured data. Recommended to move to controlled pilot in a single cost center.
Aieval qualifies and recommends — the final decision is always human.
Gaps
0No gap identified
Risks
1- Systematic errors on rare classes may go undetected
Recommendations
2- Keep human-in-the-loop on exceptions for the first 90 days
- Monitor drift monthly in SAP
Discovery
Answers collected during intake
- Problem
- Team spends 120h/month manually classifying invoices.
- Objective
- Automate 90% of classification with exception-based review.
- Impacted users
- Accounts payable team (12 people)
- Vendor / model
- AWS Textract + internal classifier · OCR + supervised classification
- Data used
- Invoices (PDF), Chart of accounts
- Integrations
- SAP
- KPIs
- % automatic classification · Post-review errors · Hours saved
- Estimated cost
- $45,000
- Risk flags
- Uses PII: NoCustomer-facing: NoProduction access: YesExternal sharing: No