The attempt usually looks the same. You download on-hand from the 3PL portal, open the inventory report the retailer sends, pull last month of shipments out of NetSuite, and export direct-to-consumer orders from Shopify. Then you paste all four into Claude and ask what needs reordering before the end of the month.
For one product in one location, that can work, and the explanation may be more useful than the spreadsheet formula it replaces. The workflow breaks when the position spans sources that use different product codes, pack configurations, locations, and refresh times. This article separates the reasoning AI handles well from the data work that must persist outside a chat.
Where AI inventory planning works today
Give a model one file and a clear question and it holds up well. It will read an export with no header documentation and tell you which column is on-hand and which is committed, which saves the twenty minutes you would have spent asking the 3PL. It will write the coverage math, explain why a safety stock formula produced that number, and show the arithmetic when you challenge it.
It is also good at the reasoning around the number. Ask why a SKU with eight weeks of cover is still stocking out at one account and a model will work through the plausible causes: a promotion pulling volume forward, stock sitting at a location that account does not draw from, a lead time that no longer matches what the supplier quotes. Those are the questions a planner would ask, and having them listed in ten seconds is worth having.
Where the whole picture fits in one place, this is enough. A brand selling one line through Shopify and one 3PL can run a planning cycle out of a chat window and a spreadsheet, and should, until the third channel arrives.
Where it breaks: the position lives in four places
Right answers are hard to pull out of large datasets, and inventory is a large dataset wearing the disguise of a small one. The number you want, units of a product available to sell against demand you can see, is assembled from sources that each hold a fraction of it and disagree about what a product is.
| Source | What it knows | What it cannot tell you |
|---|---|---|
| 3PL or warehouse portal | On-hand and committed by location, in cases | What is already sold through at retail |
| Retailer inventory report | Store and DC stock for their doors, by their item number | Anything about your other accounts |
| Your ERP | Shipments, open orders, standard cost | Whether the shipment sold or is sitting in a back room |
| DTC platform | Units sold and on-hand for the channel you control | The wholesale half of demand |
Pasting all four exports creates three risks. A manual paste can omit a sheet, filter, or location, and the model cannot identify data it never received. Product mappings explained in one session do not become durable business rules, so a 12-count and a 12-pack can be merged on name similarity. The next session may also receive newer files without a record of what changed, which makes two different answers impossible to reconcile.
- Rows dropped without an error
- Mapping re-explained every week
- Different answer, no change log
- Every location in the total
- One product key, written once
- Same answer, or a logged change
How Permute makes the position durable
Permute is the context layer we sell between inventory sources and the AI and planning tools that use them. We connect the 3PL feed, retailer report, NetSuite, and Shopify, then resolve their identifiers into one product record across every location.
On-hand, committed, in transit, and sell-through are defined once and applied wherever they are used. Pack conversions carry effective dates, and each source retains its refresh time, so a late 3PL file is shown as stale rather than presented as a complete current position.
A planner can ask which SKUs fall under four weeks of cover at the current run rate and receive the answer with its contributing rows. We sit underneath Claude, ChatGPT, and Copilot, giving those assistants governed data and definitions rather than replacing their reasoning. What a context layer actually does explains the mechanism.
See your stock position in one place
Connect a 3PL feed, your ERP and your DTC channel, then ask what needs reordering.
Where Permute stops
We are not a forecasting engine. We supply the governed position and sales history that a forecast uses, while the projection stays in a planning tool or model. We also read from connected systems without writing back, so a person still approves and raises the purchase order in the ERP.
Inventory decisions that stay human
- The buy itself. Cash, minimum order quantities and what you are willing to be wrong about are a judgment call, and the person who owns the outcome should make it.
- The exception. A planner decides whether a late feed makes the position too stale to use and whether a supplier commitment can wait for the correction.
- The product rule. A person who knows the catalog approves whether a replacement SKU inherits the history of the discontinued item.
Those decisions become reviewable when their inputs identify every included location and the time each source last refreshed.
A test worth running this week
Take one SKU that stocked out in the last quarter (the one everybody still brings up) and reconstruct the week before it happened. Write down the on-hand you would have seen in each source, the lead time you would have used, and the demand signal available at the time. Then ask what it would have taken to see it coming.
Compare the answer with what the team could see at the time, but do not assume the finding. The test should reveal whether the miss came from source freshness, a pack conversion, an omitted location, the forecast, or the decision itself.
Use AI after the position is defined
General-purpose AI is useful for explaining formulas, testing assumptions, and surfacing questions a planner should ask. It becomes unreliable when each session has to reconstruct product identity, source coverage, and freshness from pasted exports. Preserve those facts outside the chat, then let the model reason over them. The planning decision remains human, but the evidence behind it no longer has to be rebuilt each week.
Bring your own exports to the call
We will walk through where the position fragments across your warehouses, retailers and 3PLs, and what it takes to close the gap.
Questions ops teams ask about AI and inventory
What should happen when an inventory feed is late?
Keep the last successful materialization available, label it with its refresh time, and identify the missing source. The planner can then decide whether the position is current enough for the decision instead of receiving an incomplete total presented as current.
How much history does this need before it is useful?
Current-position questions can be useful with recent on-hand, open-order, and sales data. Forecasting needs enough history to represent the seasonal and promotional patterns relevant to the product, so the required period depends on the business rather than a universal month count.
Can an outside broker or agency use the same inventory view?
Yes, if access is scoped to the accounts, products, and fields that party is entitled to see. The governed position should not require copying a broader workbook and deleting tabs before every share.
How should an unresolved pack conversion appear?
Leave the affected rows unresolved and exclude them from totals that require a common unit. Show which product and source need a decision, because an explicit gap is safer than a plausible conversion applied silently.