Small and medium-sized businesses can benefit from generative AI that is intelligently incorporated into a supply chain application without having to pay a fortune, hire expensive IT talent, or experiment with the bleeding-edge technology in the hope of finding a use case with good ROI.

Over the last two days, I have been briefed by a couple of Netstock executives, seen a product demo, and spoken with a Netstock customer. Netstock , with revenues above $50 million, provides an economically priced public cloud-based inventory planning solution to small and medium-sized businesses.

Let’s start with the customer. I interviewed Jared Ramkellowan, a supply chain and IT manager at CC1 Sint Maarten. Americans may be more familiar with the name “St. Martin.” St. Martin is on the northern French side of the island, while Sint Maarten is on the southern Dutch side.

CC1 Sint Maarten Optimizes Inventory and Improves Service with Netstock

CC1 Sint Maarten is one of the CC1 Companies. This conglomerate is a major beverage manufacturer, bottler, and distributor headquartered in Puerto Rico, serving the Caribbean. CC1 distributes coffee, beer, spirits, and bottles and distributes Coca-Cola products.

CC1 Sint Maarten imports everything from 200 suppliers located mostly in the Caribbean. There are 1,200 stock keeping units .

Mr. Ramkellowan felt the company had strong supply chain expertise, but it is always looking to improve. Moving to Netstock could help them eliminate overly manual, Excel-driven supply chain operations and improve their core sales and operations planning process.

And despite being a mid-sized company, the company invests in IT – like ERP and BI - and is comfortable with it. “This was the most natural next step,” Ramkellowan said. The planning tool would allow them to “spend less time developing the signal (and) more time executing, strategizing around the signal.” A demand signal is a near‑real‑time data indicator—such as POS sales, orders, or inventory withdrawals—that reflects actual customer demand.

Sint Maarten has used Sage as their ERP solution for the last 8 years. “We really wanted something that was quick time to value and quick time to implementation.” Because Netstock was a Sage partner, it was pre-integrated with the ERP system. Additionally, it looked like an easy-to-use solution that the sales supervisor and the marketing team could use without needing to be supply chain experts. Because marketing and sales could see what was happening, buyers and planners had extra incentive to pay close attention to their actions.

The solution was easy to implement. While the whole implementation was 12 weeks – very fast for a supply chain planning application – they were connected to Sage within the first few days.

The distributor was able to implement quickly because they came into the project with clean data – an accurate item master, lead times, minimum order quantities, and so forth.

“We started seeing live data that made sense immediately,” Ramkellowan said. But to overcome cultural resistance to the new solution, the existing demand and inventory planning process was run in parallel. A planner could review the Netstock recommendations and compare them with their SKU plans. When the plans differed, they could tunnel into Netstock to understand why it produced the recommendations it did. The way we sold it to our employees was to say, “This is a tool. Let's challenge it!”

“So, my main planner, the brand managers would come in and say, ‘Okay, our Excel sheet says 1000 cases, Netstock says 1500. Why?” Is there context that Netstock doesn't know? Is there tribal knowledge that only the planner knows? Is there a promotion or a campaign that's happening that Netstock isn't aware of? We were able to fine-tune it and get to the point where Netstock (and the old process) were aligned. Netstock's numbers, recommendations, and opportunities made sense.”

All key parameters were in place, and inventory recommendations were now grounded in statistical rigor. At that point, it became a much faster, less manual process.

CC1 Sint Maarten has been using Netstock for 8 months. It has reduced its inventory while improving service using Netstock. Inventory levels have been reduced by 20%, inventory turns increased by 20%, and excess inventory is down 10%. Meanwhile, the inventory planners are 20-30% more productive.

On the customer service side, stock-outs have been reduced by 50%. “The ultimate goal for us is ensuring our customers have the best experience,” Ramkellowan explained. They can use visibility to plan more effectively and make informed trade-offs. If they will be out of an item for two days, what other items can they promote to keep sales up? Or if a promotion is planned, and fulfilling the plan will be difficult, should the promotion be canceled? If they have excess inventory, should prices be cut to get those SKUs out of the warehouse? “That is the strategic part.” Their S&OP process continues to improve with the new solution, and sales personnel now have the information on hand to explain stockouts to customers.

AI did not drive CC1’s decision to implement Netstock. “AI is a tool that allows us to move faster and get better. But I think the main thing about the AI is that the decision accountability stays with the people,” Ramkellowan concluded.

The core of any demand and inventory planning tool is statistical modeling and machine learning. It is not generative AI. At Netstock, because it is a public cloud solution, data from all customers can be aggregated to develop better industry-specific forecasting algorithms. This is done while protecting the sanctity of an individual customer’s data.

Generative AI does, however, improve the solution by enhancing the explainability of the inventory suggestions. Netstock has developed its own large language models leveraging its data mart. Because this is not based on public data, it is more accurate.

When a user opens Netstock, they first see the landing dashboard. Netstock is undertaking an inventory calculation for every part number and every SKU at every planning location the organization has. These calculations are driven by a target fill rate, the percent of expected demand the company wants to fill. And that number can be driven by how much cash the company is willing to spend on inventory to hit a fill rate target.

The landing dashboard directs a planner to focus on the most important customer service threats, such as stockouts, and key opportunities, such as running a promo to reduce excess inventory. These are not just current opportunities and threats. The dashboard looks ahead and surfaces issues. This allows planners to mitigate or work through potential issues before they become reality. None of this requires generative AI.

But if a planner sees a recommendation and wants to understand what is driving it, instead of drilling into tabular data and calculations, the AI can explain what is driving the recommendation. For example, in the demo, one AI recommendation read “Product BQ5103, Axle Beam and Brakes ROR Steer Axle > in Chicago Central Warehouse has 20 units of surplus orders. The products you ordered will cost $102.60 and you don’t need them. Delay those orders, or even better, cancel them as soon as possible.” This speeds decision-making and allows planners to handle more issues.

Netstock is also working towards agentic AI capabilities, the ability for the engine to kick off actions, like an email to a supplier canceling an order. This will be delivered in the first quarter of next year.

But while they develop these capabilities, Netstock does so with some trepidation. They firmly believe in planner-in-the-loop decision-making.

Justin Dudley, a solutions consultant at Netstock, gave this example. Imagine the agent sees excess inventory being ordered and automatically reduces the order? But what if the excess inventory is the result of executives deciding that, because of new tariffs, they are going to front-load ordering and save on duties? Until generative AI can fully understand the context of a decision, automatic decision-making should proceed cautiously.