- Industry: Manufacturing
- Platform: Cassidy
- Challenge: Turning ERP reports into timely purchasing and production decisions
- Result: AI workflows reduced manual purchasing checks and surfaced 40+ new use cases and workflow refinements in a single day of onsite training.
The Challenge: Finding the Signal in ERP Data
DEMA Engineering Company manufactures chemical dispensing systems and industrial valves. Its supply chain team manages purchasing decisions across products with different usage patterns, supplier terms, and lead times.
Epicor, the client’s ERP, holds usage history, inventory levels, supplier pricing, and buying suggestions generated from production needs. The team’s bottleneck was turning that data into a decision fast enough to act on.
Before placing an order, buyers had to check whether an item already had an active supplier price agreement or needed a fresh quote, a comparison that meant manually cross-referencing the buying suggestions report against the supplier price list for every item coming due. On top of that, they worked through a steady stream of Epicor status messages: quantity changes, schedule adjustments, releases, cancellations.
Parts sourced from an international supplier carried more risk than the rest: roughly 100 items on a nine-month lead time, where a late decision could create a production gap that took months to close.
The company had standardized on Cassidy as its AI platform, but it couldn't connect directly to Epicor. Each workflow started with someone manually exporting a report. David Yuill, Director of Supply Chain Operations, didn’t wait on the integration. He started turning those reports into workflows his team could act on.
The Solution: Turning Operational Rules Into Workflows
SoftSnow initially worked with the client to build two purchasing workflows.
The first reads a past-due order export and prepares supplier-specific email drafts for buyers to review, replacing the vendor emails buyers had been composing by hand roughly twice a week.
The second focused on inventory optimization across roughly 1,000 products. SoftSnow and the operations team scoped a model covering economic order quantity, safety stock, and reorder points, giving buyers a single structured output instead of three separate reports to reconcile by hand.
Across those two projects, SoftSnow walked DEMA's supply chain leader through the principle behind every workflow it builds: start with the input, add the right context, define a clear output. That became the template for what he built next:
- Start with an existing Epicor report.
- Export the data into Excel.
- Encode the team's rules and thresholds.
- Produce a structured output employees can act on.
Working through those first two builds alongside SoftSnow gave him a model he trusted enough to keep using on his own.
When automating their ERP exports became a constraint, the supply chain leader continued developing the inventory concept himself. He applied it to one of the company's highest-risk supplier relationships, where the nine-month lead time made earlier visibility especially important.
The resulting report helps identify changes in usage and highlights items that may require action. That same order-quantity thinking also appears in another workflow he built: a cross-reference between supplier price lists and Epicor's buying suggestions, the same manual check buyers used to do by hand, so a fresh quote only goes out when one is actually needed. It also factors annual usage into supplier price breaks and recommends purchasing quantities.
David then extended the same method beyond purchasing. He built two daily production exception workflows: one flags production records where the employee associated with a transaction does not match the assigned work center, helping the team identify possible reporting errors; the other identifies production efficiency above 125 percent or below 75 percent.
Together, the checks cover about 85 percent of what the team wants to identify, with further refinements still underway.
As David noted, "If your data entry error was within the guidelines, it would not catch it." That limitation gives the team a clear view of where human review and further refinement are still needed.
The Result: One Builder Creates a Repeatable Model
The AI workflows gave DEMA's team faster answers to the exact problems that started this story: fewer manual cross-references before a purchase order, earlier visibility on the supplier relationship with the longest lead time, and a daily check on production data that used to depend entirely on someone catching it by eye.
During onsite department sessions, the SoftSnow team used existing workflows, including David's own builds, as teaching material. Employees saw how reports they already worked with every day could become inputs for structured AI processes. More than 40 new use cases and workflow refinements surfaced during a single day of training, generated by the participants themselves.
The clearest proof came from leadership. Asked to describe the return on DEMA's AI investment, President Jeremy Deutsch pointed to David's international supplier report: "Small thing but could be very impactful." He had already described the same example to a Fortune 500 manufacturing partner, using it to show how quickly a mid-size manufacturer could put AI into practice.
DEMA's growing independence didn't reduce how much they leaned on SoftSnow. It generated more of it: the more comfortable David and his colleagues got building on their own, the more new problems they found worth bringing back to build through together.
Ready to turn the reports your team already uses into practical AI workflows? Let's talk.
