Operations Teams Guide

Purpose: Extract structured data from invoice documents for automated entry and reconciliation

Inputs:

• Invoice document (PDF or image)
• Known vendor list for validation
• Purchase order references
• Required field specifications

Task Instructions:

Extract structured data from {invoice_document}. Identify vendor information, 
invoice metadata, line items with descriptions and amounts, and calculate totals. 
Validate against {vendor_list} and cross-reference with {purchase_order} if 
available. Return extraction confidence scores for each field.

Outputs:

• Vendor details (name, address, tax ID)
• Invoice metadata (number, date, terms)
• Line items with descriptions and amounts
• Totals breakdown (subtotal, tax, total)
• Extraction confidence scores

Integration: Document upload to storage system or email receipt, create ERP entry, flag exceptions for review

Purpose: Identify key terms, obligations, and risk clauses in legal contracts

Inputs:

• Contract document (PDF)
• Contract type (e.g., NDA, MSA, SOW)
• Required clause checklist
• Risk keywords list

Task Instructions:

Analyse {contract_document} and extract key terms. Identify all parties, 
contract duration, financial terms, obligations, and liability clauses. 
Flag any terms in {risk_keywords} and verify presence of {required_clauses}. 
Extract renewal dates and termination conditions.

Outputs:

• Party identification
• Term duration and renewal dates
• Financial obligations
• Key clauses and obligations
• Risk flags and missing clauses

Integration: Contract uploaded to document management system, update contract database, alert legal team of risks

Purpose: Extract and validate data from applications, submissions, and standardised forms

Inputs:

• Form document (PDF, image, or scanned)
• Form template specification
• Validation rules
• Required fields list

Task Instructions:

Extract all field values from {form_document} based on {form_template}. 
Normalise data formats, validate against {validation_rules}, check completeness 
against {required_fields}, and assign a completeness score. Flag any fields 
that fail validation or are illegible.

Outputs:

• Extracted field values
• Data normalised to standard formats
• Completeness score (0-100)
• Validation failures
• Fields requiring manual review

Integration: Form submission via email, portal, or scan, create record in database, route to review queue if incomplete