AI Automation

Supplier invoices out of the inbox and into the ledger

Three staff spent most of each morning rekeying supplier invoices from PDFs and email bodies into the accounting system.

Client
Building supplies wholesaler
Sector
Distribution
Year
2026
Service
AI Automation
~22 hrs
Manual work removed weekly
94%
Straight-through processing
100%
Decisions logged for audit

Engagement

Stack involved

  • Claude API
  • Python
  • Power Automate
  • Microsoft Graph
  • SQL Server

The situation

Around 400 supplier invoices arrived weekly across PDF attachments, scanned images and plain email text, in no consistent format. Rekeying was slow and error-prone, and errors surfaced at month-end when they were expensive to unpick.

What we did

  1. 01

    Two weeks measuring the existing process to establish a real time-and-error baseline before proposing anything.

  2. 02

    Extraction pipeline for line items, totals, tax and purchase-order references, running on an enterprise API tier with no training on client data.

  3. 03

    Confidence thresholds tuned so uncertain extractions route to a person with the source document and the model’s reasoning attached.

  4. 04

    Three-way match against purchase orders and goods-received notes before anything posts to the ledger.

  5. 05

    Every decision logged with input, output and confidence, giving the auditors a trail that did not previously exist.

The outcome

Ninety-four percent of invoices now post without human touch; the remainder queue for review with the context needed to clear them quickly. Roughly 22 hours of weekly rekeying was removed, and the two staff who did it moved onto supplier queries and credit control.

Service used AI Automation

Tell us what is breaking.

A short call, an honest read on whether we are the right fit, and a fixed-price proposal after discovery. No obligation and no sales sequence afterwards.

Book a consultation hello@queuebytes.com

Mon–Fri, 09:00–18:00 · Emergency cover 24/7