How Mercos cut 70% of accounts payable work with invoice processing automation (and proved finance can build too)

“Wait… why are we still doing this manually?”

That was the question Jonattan Fenrich, FP&A Lead at Mercos, kept running into. Hundreds of invoices, all landing at once. Each one needing to be opened, read, classified, and keyed into the ERP. By hand. Every month.

For most mid-sized finance teams, month-end close still means long hours, spreadsheets, and inboxes full of PDFs. At Mercos, the team asked a simple question: what if we didn’t have to open every single invoice? That’s what led them to rethink their process, and rebuild it with accounts payable automation.

Here’s the story of how an overwhelmed finance team turned into a scalable finance operations engine.

Meet Mercos: The team behind the change

Carlos Marian, CFO, and Jonattan Fenrich, FP&A Specialist, have been leading the charge. Their goal? Give analysts the tools to automate manual work and focus on what really matters: control, cash flow, and insights.

"We're not trying to scale by adding more people, we're trying to scale by getting smarter. That means automation at the core of finance." — Carlos Marian, CFO

Why now: The moment that broke the process

Every month-end, things piled up:

• ~400 invoices
• Dozens of manual forms
• One analyst working late

Jonattan described it simply: “It was work with a lot of clearly automatable tasks.” Documents arrived by email. Each had to be opened, read, renamed, entered into Granatum, and saved into Google Drive.

• Most of the documents received/entered were invoices (NFs).
• There are other types of documents, such as bills and tax guides, but they represent a smaller share of the total volume.

Instead, they decided to prototype something better, and prove automation could handle the load.

The automation: What we built together

They started with AP, the biggest source of manual work.

Mercos didn’t want another tool to manage. They wanted to build their own AP automation tool, something flexible, fast, and finance-led.

So they used Abstra to build one core flow for invoice processing.

Here’s how it works:

*Some invoices (NFs), even when posted automatically, require manual completion. For example, occasional expenses (where no standard is registered in the allocation table) and cost splits (which depend on specific criteria).

• 📥 Input: Incoming invoices (PDFs) received via Gmail
• 🤖 Abstra automates:
  • Extracting attachments
  • Validating if the file is an actual invoice
  • Extracting fields like Tax ID, amount, and issue date
  • Launching data into Granatum ERP
  • Sending alerts via Slack
• 📤 Output: Clean, structured invoice data and documents, automatically synced with ERP and ready for AP review

That’s it. This end-to-end workflow is a real example of automated invoice entry in action. Jonattan built and tested everything himself, no devs needed.

"This wasn't plug-and-play. It was thoughtful design. And it worked."

It took just under 2 months to get from idea to production. That included prototyping, validation, and migrating the full flow into production using Abstra.

AI played a key role throughout this flow. Abstra uses AI models to read and extract invoice data like Tax ID, value, and issue date, removing the need for templated OCR or manual mapping. But more than that, AI also supports the finance team during development.

Jonattan used Abstra’s hybrid builder, which combines natural language prompts, drag-and-drop logic, and Python, to write and test code faster, without being a developer. That means faster prototypes, fewer bugs, and a better feedback loop.