Case Study: Logistics Industry

Automating Purchase Invoice Matching

From four people matching invoices by hand to one person reviewing the exceptions

By Lana Korzhuk, Founder & CEO

At a Glance

Hundreds of purchase invoices arrived each week, and four people spent much of their time matching them against goods-receipt records. We built a workflow that completes routine matches, flags anything uncertain, and leaves the team with a short exceptions queue instead of a full invoice backlog.

Challenges: High-Volume, Manual Invoice Processing

The team: a high-volume logistics operation handling different invoice rules across suppliers and depots.

The client handled 400-500 purchase invoices each week, each requiring manual entry and verification against detailed goods receipt notes (GRNs). This process presented several critical challenges:

Before and After Automation Metrics
Metric	Before Automation	After Automation
Staff Time	Project baseline	75% saved
Manual Processing	Project baseline	65% lower within 90 days
Annual Saving	Project baseline	Approximately £80,000
Error Rate	Frequent Duplications, incorrect values	Minimised, flagged clearly

Resource-Intensive:

Manual invoice processing consumed substantial staff capacity across the finance workflow.

Error-Prone:

Duplicate entries and manual errors were frequent, impacting data accuracy.

Time-Consuming:

Identifying and correcting errors took significant time, delaying the process and increasing operational costs.

Solution: Tailored Automation with Microsoft Power Platform and SQL Server

The solution was designed to streamline the entire purchase invoice process, incorporating tailored automation tools that adapted to the client's unique business rules. The approach consisted of three core stages:

Invoice Ingestion (Power Automate)

Using Microsoft Power Automate, incoming invoices were automatically routed to an AI processing hub (AI Builder by Microsoft) where they were standardised and converted into structured text files.

Standardisation and Matching (AI Builder and SQL Server)

Processed invoices were transferred to the client's server, where SQL Server procedures matched them against corresponding GRNs based on unique criteria like invoice number, supplier reference, and values. The solution adapted to specific matching rules for each supplier and handled invoices with multi-GRN and depot-specific ledger postings.

Auto-authorization and Error Flagging

Custom rules were set to automatically authorise invoices within accepted variances and flag discrepancies. This allowed the system to auto-approve or hold invoices based on predetermined conditions, significantly reducing manual intervention.

Technologies Used:

Microsoft Power Automate
Microsoft AI Builder
SQL Server (Procedures and Agent Jobs)

Results: Enhanced Efficiency, Cost Savings, and Accuracy

Key Outcomes

Staff Time Saved
The invoice-matching workflow saved 75% of the staff time previously used on the process.
Manual Processing Reduced
Manual processing reduced by 65% within 90 days.
Measured Annual Saving
The resulting annual saving was approximately £80,000 for this implementation.
Operational Cost Savings
With reduced labour needs and faster processing times, the client experienced a substantial decrease in operational costs.

Processing Breakdown

Manual Processing Reduced (65%)

Remaining Baseline (35%)

Manual Review Reasons:

GRN not recorded in system
Quantity mismatch beyond threshold
Price variance exceeding allowed limit
New supplier or product requiring setup

Evidence note

The client remains anonymous. The three figures above all belong to this invoice-matching programme; no quotation is attributed without approval.

Explore the service behind this result: AI document processing and workflow automation.

Frequently Asked Questions

What results are approved for publication?▾

The approved figures are 75% of staff time saved, a 65% reduction in manual processing within 90 days and an approximately £80,000 annual saving. All three belong to this invoice-matching implementation.

How long does invoice automation take to show results?▾

In this project, manual processing reduced by 65% within 90 days. The work was in encoding supplier matching rules, validation and exception handling rather than waiting for a model to learn without supervision.

What technology was used to automate the invoice processing?▾

Microsoft Power Automate handled invoice ingestion and routing, Microsoft AI Builder extracted and standardised the invoice data, and SQL Server stored procedures matched invoices against goods receipt notes using each supplier's own rules. No rip-and-replace of the client's existing finance system was required.

Does automated invoice matching handle multi-GRN and depot-specific invoices?▾

Yes. The matching logic accounted for invoices spanning multiple goods receipt notes and for depot-specific ledger postings, because those were the cases that made the manual process slow in the first place.

What happens when an invoice does not match?▾

Custom rules auto-authorise invoices that fall within accepted variances and hold anything outside them. Held invoices are flagged with the specific discrepancy, so the reviewer sees why it stopped rather than re-doing the comparison themselves.

Would this work for a business that is not in logistics?▾

The pattern — extract, standardise, match against a source document, auto-approve within tolerance — applies to any business processing a high volume of purchase invoices against receipts or purchase orders. The supplier-specific matching rules are what get rebuilt for each client.

Still Matching Invoices by Hand?

Show us how invoices move through your business today. We'll tell you where automation could help and where a person still needs to stay involved.

Request Workflow Review