Fluent for Epicor Kinetic: “Everyone said it was too good to be true. It works.”
Sixteen years into Epicor, Metalworks handed its AP invoices to Fluent’s AI agents. It connected in a day, 99.4% of what it read was right, and an approval that used to take five minutes now takes seconds.


Meet Jill Schoedel
Jill Schoedel is the IT Manager at Metalworks in Ludington, Michigan, and she did not arrive at the top of the org chart. She climbed it. “I’ve kind of worked my way up through the ranks over the years as IT manager,” she says, which undersells being the person the plant floor and the front office both call.
Her Epicor story is nearly as long. “We’ve been with Epicor since, I believe it’s September 2010 or 2011.” In 2022 she moved the whole operation to the cloud. And every couple of years she makes the trip to the Insights Epicor Global Customer Conference.
A Booth Visit at the End of the Conference
AI was one of Metalworks’ main initiatives this year, so when the invitation to stop by the Technology Coast Partners booth, where they were showing Fluent, arrived, Jill and Guy Wolgamott, Metalworks’ application analyst, went. Near the end of Insights, they took a quick demo. “We both thought, this is really great. We needed to take this back to Metalworks.” Back home, the reaction was the one every AI pitch gets.
"Everyone was saying that it was too good to be true. But actually, it proved that it is a good software, and that people should not feel like it’s too good to be true, because it works."
It Connected to Epicor Right Away, then it Just Worked
There was no implementation project and no issues. Andres Cubeddu, Head of Customer Success at Technology Coast Partners, walked Jill through the API key and the access scope, then worked with Metalworks’ AP specialist and Parker Kerwin to get the invoice flows running the right way.
The Numbers Behind the Accuracy
Metalworks’ own Fluent document analytics. Field retention of 99.6% across 298 scored AP documents. When reviewers did open a draft, they left the AI’s extraction exactly as it was. The supporting chart, titled “Document Analytics Review — Field Retention,” carried the headline. “When reviewers made edits, they changed almost nothing,” illustrated with a 99.6% circular gauge, and explained: field retention was 99.6% across 298 scored documents, meaning the AI's original extraction was almost always kept exactly as-is — the clearest sign that Fluent is reading the documents correctly. In practice, reviewer edits typically just added internal details like dates, references, or accounting information that don't appear on the source invoice. The takeaway called out on the chart: across 298 documents, the AI extracted 99.6% of fields correctly, with no human correction needed.
Overall performance broke down as:
The People who Touch it Every Day Stopped Shuffling Paper
Before Fluent, an AP invoice was verified, keyed into Epicor and balanced by hand: two minutes to fifteen, call it seven or eight on average. Approvals meant writing okay to pay on paper, copying it and walking it over.
The Before and After Fluent
Entering one AP invoice: 7–8 minutes before Fluent → under a minute to review with Fluent
Getting an invoice approved: 5 minutes at best before Fluent → seconds, from an email, with Fluent
Preliminary projected annual savings: $35,700 a year.
"There wasn’t really any issues. Working with Andres, he made it very simple for us. He just made it super easy, and we can’t thank him enough."
End-User Acceptance
The approvals agent sends an email with a button. “From what I've heard from the end users, they love it. They look at it, and if everything's good, they just click approve.”
In AP, the source invoice sits beside the draft to be checked before anything posts. “It just makes our accounts payable team's day a lot easier. There's not so much paperwork to shuffle.”
Compared with the Last AI They Bought
Metalworks had spent about a year fighting another AI tool on its quoting process. “I just see errors every other week. With Fluent, we're not seeing that at all.”
Her word for the difference: “It's like a breath of fresh air. We're very, very grateful, and very much impressed.”
Why It Stuck
AI was on the agenda at the annual strategic plan. “Finding a reliable company was at the top of the list, and they are very happy with what we've implemented.”
Jill Schoedel added: “We tell Andres all the time that Fluent is just unbelievable. To see it connect to Epicor right away and start working for us during testing, we were very, very, very happy.”
What's Next
Order entry is the next goal, with accounts receivable behind it, then other offices at Metalworks, process by process.
Jill’s advice to everyone weighing it up:
“Just give it a try.”
Send us a batch of your own invoices: tcpamericas.com - fluenterp.com

Headquarters
Ludington, Michigan, USA
Founded
1966
Company size
51-200
Industry
Manufacturing
Website
www.metalworks1.com/SOLUTIONS


RESULTS
99.4% extraction accuracy over the trial
99.6% of fields kept with no correction
< 1 min to fix a document that needed a touch
$35,700: Preliminary projected annual savings


