“We weren't just slow. We were one bad week away from the whole thing falling apart. Every order required someone to sit down, open a file, read it, type it in, and pray they got it right. That's not a process. That's a liability.”
David Mercer
Director of Operations, EVO Systems
Challenge
EVO Systems had built a high-volume operation on a foundation that was never designed to hold it.
Over seven years, EVO Systems had built deep expertise in window and door components, grown their catalog to more than 25,000 SKUs, and invested in Oracle NetSuite as the backbone of their operations. What they had not built was a way to get order data into that system without a human being doing it by hand.
Three customer service representatives managed the entire intake workflow. Each morning, purchase orders arrived via email as PDF or Word attachments, each formatted differently depending on the customer, each carrying its own structure, its own part numbers, and its own quirks. A representative would open the file, read through it, cross-reference item codes, verify quantities, check shipping details, and manually key every line into NetSuite. That process took between 15 and 30 minutes per order. Across 20 to 30 orders a day, it consumed the team's entire capacity before noon.
The problem was not effort. The team worked hard. The problem was that the process had no tolerance for growth. Every new order was another 30 minutes of human attention, which meant that scaling volume meant scaling headcount, and scaling headcount meant scaling cost. EVO was growing. The system was not growing with it.
What made this harder was the variety of order types the team had to manage. Regular orders, Inventory Protection Program orders, release plans, sales orders, and orders from quotes each carried different validation requirements and different downstream paths in NetSuite. A single missed field or transposed item number did not just cause a delay. It created a chain of corrections that slowed fulfillment, frustrated customers, and pulled the same representatives away from new incoming orders to fix old ones.
We kept thinking we could hire our way out of it. Add another rep, train them up, distribute the load. But every time volume jumped, we were right back in the same position. The work was outpacing the team no matter what we did.
Solution
Cogent Labs rebuilt EVO's order intake from the ground up, replacing a manual workflow with an intelligent system that reads, validates, and submits orders without human intervention.
When Cogent Labs began working with EVO Systems, the first decision was not about technology. It was about where in the process the friction actually lived. Manual data entry was the symptom. The root cause was that there was no structured layer between an email attachment and NetSuite. Every order arrived as unstructured content, and a human being was doing all the translation work in between.
From the inbox, the order moves to the parsing engine. This engine reads the attached file, whether it arrives as a PDF, an Excel sheet, or a Word document, and extracts every relevant field: the PO number, item codes, quantities, pricing, shipping address, and any associated notes. Because EVO's customers each format their orders differently, the engine was built to handle variability rather than expect uniformity. It does not rely on the file looking a certain way. It reads meaning from structure, not from layout.
Once extracted, the data passes through an automated validation layer before it ever touches NetSuite. Item codes are checked against EVO's catalog. Customer mappings are verified. PO numbers are cross-referenced. If something does not match, the system flags it for review rather than submitting incomplete data downstream. Administrators can review flagged orders, make corrections, and submit them with confidence. Critically, every correction feeds back into the system, allowing it to recognize similar patterns in future orders and handle them without flagging.
Seeing it run for the first time was one of those moments where you just go quiet. An order came in, and by the time I finished my coffee it was already in NetSuite, validated, formatted correctly. The whole thing. Nobody touched it.
Results
EVO Systems no longer trades time for volume, and their customer service team finally has the capacity to work on what actually requires their judgment.
The numbers tell a clear story. Orders that once took 30 minutes to process now complete in under one minute. Manual data entry, which previously consumed most of the team's day, has been eliminated entirely. Data accuracy, measured against the error rates that had historically driven correction cycles and fulfillment delays, improved to 97%. EVO can now handle higher order volumes without any increase in headcount.
But what matters more than the metrics is what they made possible. Before Project Parse, the customer service team's time was largely spent on work that did not require their expertise. Reading files, typing numbers, checking boxes, fixing typos. That work is gone. The team now focuses on customer relationships, escalations, and the judgment calls that actually benefit from a human being paying attention.
For a distributor operating at EVO's scale, the ability to absorb volume without absorbing cost changes how the business can grow. New customers, seasonal peaks, and expanding SKU coverage no longer require staffing conversations before they can be pursued.
100%
reduction in manual data entry
30x
faster order processing
97%
improvement in data accuracy
We used to think about growth in terms of how many people we'd need to hire. Now we think about it differently. The system handles the volume. The team handles the customers. That's how it should have worked all along.
Conclusion
The operation had everything it needed except a system capable of closing the loop.
EVO Systems had invested in the right tools. NetSuite was mature, well-configured, and capable of running a sophisticated distribution operation. Their team was experienced. Their catalog was deep. What they were missing was not capability on either side of the workflow. It was the connection between them. Every order that arrived as an unstructured file and had to be read by a human being before NetSuite could see it was a point where speed, accuracy, and scale were being surrendered by default.
The pattern is not unique to distribution. It appears in any business where valuable structured systems sit downstream of unstructured inputs, and where the work of translating between the two falls on people who have more important things to do. The document that needs to become a record. The email that needs to become a task. The file that needs to become a transaction. When humans are filling those gaps, the business is slower and more fragile than it needs to be.
What Project Parse demonstrated at EVO is that removing human beings from translation work does not reduce human involvement in the business. It increases the quality of that involvement. When the repetitive work is handled by a system designed for it, the people in the room can focus on the decisions that actually require them.
Have an idea worth building with AI?
Tell us about your workflow. We'll show you what a custom AI system can do.
Book a discovery call