Indian Railway - Expenditure Agent
Fabric Score 3.5
PublicFinanceIndian RailwaysNon-GenAI modelExecutive / Org Leadership ×1User / Operator / Front-line Staff ×2
Workflow Diagram
Fabric Score
Task
Perform internal financial checks on uploaded procurement documents (purchase order, receipt note, invoice, inspection certificate, etc.) and generate a finance note/return note for human review.
Intent
Improve accuracy, transparency, and consistency of bill processing, while reducing manual effort.
AI Workflow
Input
A user uploads bill-related documentation, including purchase orders, invoices, receipt notes, inspection certificates, and supporting records.
Process
The AI system performs OCR-based text extraction, applies financial and compliance rule checks, validates documentation against contract terms, and detects errors, mismatches, or missing information to determine documentation completeness.
Output
The AI system produces either a preliminary Finance Note or a Return Note, which is reviewed and finalised by a human finance officer who may accept or modify the output.
Human Oversight Level
Human-Led with AI Assistance
Institutional Oversight Examples
- •Internal finance governance
- •Procurement policies
- •Audit compliance rules
- •Digital record traceability accountability through documented notes
Risk
Incorrect model assumptions may lead to inappropriate acceptance or return notes if unchecked.
Output Modification Telemetry
- Factual errors / hallucinations100%
Transversal Metrics
Grouped by Fabric dimension.
Efficacy
Accuracy85 %
Efficiency
Modification Rate10 %
Modification Time3 min/output
Verification Time2.7 min
Rejection Rate2 %
Operational Friction5 hrs/wk
Implementation Overhead200 hrs
Governance Overhead36.7 hrs
Time to Launch12 wks
Value
Effort Reduction81.7 %
Risk
Reliability2 incident/month
Autonomy90 %