How a transport refrigeration dealer automated 32,000 vendor invoices a year

vendor invoices a year run through one automated workflow

%

vendor coverage, up from just the 3 largest under the previous tool

estimated staff hours saved per year

The Bottom Line

A Midwest transport refrigeration dealer receives about 32,000 vendor invoices a year, and each one has to become a record in its ERP before it can be paid. Its existing automation only worked for its three largest vendors, and most invoices still needed someone to fix them by hand.

ProsperSpark replaced it with an AI-powered workflow that reads invoices from any vendor, checks them against open purchase orders, and sends approved invoices straight into the ERP. The accounting team now spends its time on a quick review and approval.

Situation

The client sells and services refrigeration units for trucks, trailers, and rail. Outside vendors email invoices for parts and labor. Each one has to be entered into the ERP, tied to the right unit and repair order, and sent through for payment. At 32,000 invoices a year, that works out to more than 120 every business day.

The company already had an automation, built by another vendor. It found data by looking at fixed spots on the page. If a vendor changed its layout, even by making a logo slightly bigger, the tool lost track of fields like the unit VIN. It was set up for the company’s three biggest vendors, which covered about 70% of invoices. Even then, the client said nearly every ticket needed someone to go back and correct something.

Then the company changed its email domain, and the original vendor quoted a full rebuild to update the connection. Paying that much to keep a tool that only partly worked didn’t make sense, so the company started looking for a better option.

Before, their team was correcting almost every invoice the old tool touched and entering the rest by hand. Now an invoice from any vendor comes in, gets checked against the PO, and all someone has to do is approve it.

Solution

ProsperSpark built a workflow that reads the content of each invoice, so it doesn’t depend on where a field sits on the page. Before quoting, we built a rough prototype and tested it on sample invoices from the client’s top vendors to confirm the approach would hold up.

Here’s how the finished workflow runs:

1. A vendor emails an invoice to a dedicated inbox, which starts the automation in Make.

2. Google’s AI vision tools extract the text and details from the invoice.

3. An OpenAI model interprets those details and sorts them into the right fields, like parts, labor, unit information, and amounts.

4. The invoice data is stored in a Google BigQuery database built for this process.

5. The automation pulls the list of PO numbers from the client’s ERP and checks that each invoice has a matching PO and matching amounts.

6. Invoices show up in a Softr review interface, where the team can see the extracted details and match results in one place.

7. Once a user approves an invoice, it goes into the ERP for payment automatically.

The client kept its existing ERP. ProsperSpark connected to it, which meant working through API access with the ERP vendor and setting up secure remote access.

ProsperSpark also hosts and maintains the automation. The client pays one monthly fee that covers the automation platform, AI tools, and support, so costs stay the same from month to month and changes like a domain switch don’t mean starting over.

Results

Accounting still reviews every invoice before it’s paid, but they no longer key in the data or fix what the old tool got wrong. Before, that cleanup touched most of the 32,000 invoices coming in each year.

Invoices from smaller vendors now go through the same process as the big three, and a vendor changing its template no longer stalls anything. When something shifts in the client’s environment, ProsperSpark updates the automation as part of the monthly service.

    Let's make your tech work for you, not the other way around. Schedule a FREE no-pressure chat with us today.

    At a Glance

    Client
    Midwest transport refrigeration dealer

    Organization

    • Transportation equipment sales and service

    Business Challenge

    • Existing invoice tool broke whenever a vendor changed its layout
    • Most tickets still needed manual correction
    • A domain change would have required a full rebuild

    Market Considerations

    • Dealers and service companies receive invoices in many formats from many vendors
    • Tools that read fixed page positions break easily
    • AP errors hit margin directly

    Key Takeaways

    • AI that reads invoice content handles vendor variety far better than fixed templates
    • Building on the existing ERP avoids a costly system replacement
    • PO matching beforeapproval keeps a person in control of what gets paid
    • A managed monthly model keeps costs predictable and maintenance off the client's plate