Cut weekly inventory work from 10 hours to 15 minutes.
AI automation monitors 3PL inventory daily, pulls Walmart sell-through, and triggers replenishment workflows for an 8-figure candles brand. No new software adopted.
- 10h → 15min weekly
- 100s of SKUs covered
- 0 new tools adopted
A direct-to-Walmart candles brand was running their entire 3PL replenishment off a 10-hour weekly spreadsheet ritual. The ops lead pulled inventory reports, cross-referenced Walmart sell-through, and manually triggered replenishment runs across hundreds of SKUs. Stockouts on best-sellers. Overstock on the long tail. Tuesday mornings gone forever.
The problem in one sentence
The data needed to make replenishment decisions existed; the workflow to act on it didn't.
What we built
- Daily inventory pull. AI automation monitors 3PL inventory levels every morning, parses the export, and stores normalized state — no human touches the spreadsheet anymore.
- Sell-through fold-in. Walmart sales velocity gets pulled and joined with on-hand stock, with 3 years of sales history behind it for seasonality.
- Threshold triggers. When stock crosses a per-SKU threshold (informed by velocity, not gut), the automation drafts a replenishment run and surfaces it for review.
- Exception-only review. The ops lead now sees a single morning queue of only the SKUs that need a decision — not the full catalog.
All of this runs on top of the existing WMS and 3PL portal. No new license. No platform migration. The ops lead's daily ritual stayed in the same tools — minus 9 hours and 45 minutes of manual work.
The outcome
| Metric | Before | After |
|---|---|---|
| Weekly time on inventory mgmt | 10 hours | 15 minutes |
| SKUs actively monitored | ~40 | Hundreds |
| Stockout / overstock incidents | Frequent | Material drop |
| New software the team had to learn | n/a | None |
What we deliberately didn't do
- We didn't replace the WMS. The system of record stays the system of record.
- We didn't try to fully autonomously place replenishment orders. The human reviews the exceptions — that's the right call until the model is boringly right for a quarter.
- We didn't build a new dashboard. The ops lead lives in the same tools they always did. The automation just stops surfacing the noise.