Case study · Agentic AI · Constient product, used in client engagements

AP and AR automation with a three-tier matching engine

Invoices and remittances matched in three passes, exact, fuzzy and AI-assisted, with people reviewing only what is left.

3matching passes: exact, fuzzy, then AI-assisted
The problem

What the client needed

Invoices and remittances arrive as PDFs and emails. Matching them to orders and open items was manual, and exceptions piled up at month end.

The result

What runs now

The same engine powers our AP Automation and AR Automation products.

PostgreSQLgRPCWebSocketBusiness Central AL
What we built

The system, step by step

1

A PostgreSQL schema covering prepaid, credit and tenant customer types

2

Cash application in three tiers, exact, fuzzy and AI-assisted, with human review for the remainder

3

A remittance advice ingestion pipeline

4

A multi-model pipeline (extraction, validation, matching) streaming between services over gRPC, with live updates to the user over WebSocket

5

Microsoft Dynamics 365 Business Central integration for procure-to-pay, built as AL extensions

Regulations and standards
ZATCA e-invoicingGST e-invoicingUAE e-invoicing

Have a problem you want measured?

Tell us what you want to count, check or protect. An engineer will reply, not a sales script.