Junior accountant spending full-time hours manually processing bills and receipts from scattered sources
A mid-sized services company received vendor bills and expense receipts through multiple channels — email attachments, WhatsApp photos, Telegram messages, scanned paper documents, and PDF invoices. A dedicated junior accountant spent their entire working day downloading, categorising, and manually entering each document into the accounting system — matching invoices to purchase orders, coding expense categories, verifying amounts against bank statements, and chasing colleagues for missing receipts. The process was error-prone, consistently behind schedule, and created month-end bottlenecks that delayed financial reporting.
Multi-channel AI expense processing pipeline with automated categorisation and bank reconciliation
AITENCY built an intelligent expense processing pipeline that accepted incoming bills and receipts from all channels — email (with automatic attachment extraction), WhatsApp, Telegram, and direct file uploads. The AI extracted key data from each document regardless of format — vendor name, date, amount, VAT, line items — and automatically categorised it against the company's chart of accounts. The system cross-referenced each expense against bank transaction records to verify payments, flagged discrepancies for human review, and posted verified entries directly to the accounting system. Duplicate detection, currency handling, and multi-entity routing were built into the core processing logic.
Delivered in 5 weeks: 1 week of accounting workflow audit and chart of accounts mapping; 2 weeks of document processing AI development with OCR, data extraction, and categorisation engine; 1 week of multi-channel intake integration (email, WhatsApp, Telegram, file upload portal); 1 week of bank reconciliation module and accounting system integration with parallel-run validation against manual processing.
Results
Of routine bills and receipts processed without human intervention
Junior accountant reassigned to higher-value financial analysis work
Expenses reflected in accounting system within hours instead of days or weeks
The company's month-end close process has been shortened by 3 days. Financial reporting is now based on near-real-time data rather than retrospective batch processing. The system handles an average of 60-80 documents per day across all channels with consistent accuracy, and the categorisation model improves continuously as edge cases are resolved.