At a recent session for an industrial distributor association, we walked through three ways to put AI to work on a distributor's day. The third is the one worth its own post, because it shows where prompting stops. Here it is, start to finish.
Use Case 3: Inventory, Where Prompting Stops
Start with one seasonal, heavy-duty item. The vendor sets a minimum order quantity. The lead time is long. Demand climbs for a few months of the year, then drops. You have twelve months of sales history and one decision: how much do you order, and when?
This is a sample item with sample data from the session, so nothing here is a real SKU, branch, or customer.
Give an AI tool the twelve months of sales and the vendor terms, and ask it what to do. It finds the surge. It plans for spring. It names the customers behind the demand. That's useful work.
What it doesn't do is separate the bridge from the build, test the downside, or ask the open questions a purchasing manager needs answered before committing money, such as how much cover the next order buys against the lead time.
So we ran the same prompt on the same data, with one addition: a context card of ordering rules, in three layers.
The card doesn't state the pattern. The tool still finds it. The card adds the framework: bridge versus build, downside, upside, and the purchasing manager's questions.
With the card attached, the tool shows its work on four checks. A person confirms each one.
For your top 20 problem SKUs, prompting works. One export, one card, one chat per SKU, and you answer the purchasing manager's questions.
For your full catalog, typing doesn't scale. You can't run four gates on every SKU every week by typing prompts. That's where it goes from a project to a system.
A tool that finds the pattern isn't the same as a decision someone owns. Seasonality, vendor rules, working capital, and customer commitments each need a person to confirm them, and they need confirming on every SKU, not only the ones you thought to ask about.
The research points the same way. 74% of distributors report no measurable improvement from the inventory tools they already use (NAW/MDM, "In Pursuit of Value," April 2026, n=426). Finding a pattern was never the hard part. Running the checks consistently, across the whole catalog, is.
If you want to know what your pricing, costs, and inventory decisions look like across the full catalog rather than your top 20 SKUs, that's what our Margin Diagnostic is built to find, in one analysis.
With 30+ years in distribution and nearly a decade of developing and deploying Machine Learning models tailored specifically for distributors, he helps mid-market industrial distributors identify and eliminate margin leakage across pricing, costs, and inventory — and keep it fixed. He built Intuilize on the premise that software alone doesn't earn trust and expertise alone doesn't scale: distributors need both a model built for their business and someone who knows distribution well enough to drive adoption and deliver real ROI
Contact: nelson@intuilize.com | LinkedIn