Can AI help catch stranded and missing FBA inventory?

The problem, briefly

The usual FBA inventory trouble comes from three places. Stock is stranded, meaning Amazon holds units that have no active listing, so no customer can buy them. Stock sits in a reserved or in-process status and looks missing. Or the numbers Amazon shows do not match what you sent. Amazon's own seller guidance describes stranded inventory as stock stuck in a fulfillment center and unable to be sold, for example because the item has no active listing, and it says you still pay storage costs on it. For the problem itself, start with my guide to why Amazon FBA inventory is not updating or showing.

The painful part is timing. These problems are easy to fix in week one and expensive in month three, and nobody notices them because the dashboard does not shout. That is exactly the kind of job a tireless reader is good at. If you sell on several channels, pair this with a proper stock system, covered in my inventory management system guide.

Where AI genuinely helps with this problem

• Weekly exception reading. Download your FBA inventory report as a spreadsheet and ask the model to list every SKU (your own product code for an item) that is stranded, unfulfillable, or whose units do not add up across available, reserved and inbound columns. A person can do this, but nobody does it every week across 200 SKUs.
• Comparing Amazon to your own records. Paste Amazon's count next to your warehouse or accounting count and ask for every SKU where they differ by more than a set number. Mismatches are how missing units get found early, while the paperwork is still fresh.
• Drafting the Seller Support case. When a SKU is stuck, the case needs the ASIN (Amazon's own product number), the SKU, the shipment ID, dates and a clear ask. A model can turn a few messy notes into a clean, polite case in a minute, which saves back-and-forth.
• Spotting slow movers before they get expensive. Amazon treats stock as excess when there is over 90 days of supply compared with customer demand. Ask the model to list SKUs above that line so you can mark them down or remove them before extra fees apply.

Fee note, as of November 2026. Amazon says most of its 2027 US fee changes start on 15 January 2027, and that from 1 February 2027 one monthly storage fee, with a rate set by how old the stock is, replaces the current monthly storage fee, storage utilization surcharge and aged inventory surcharge. Amazon says stock that turns in under six months pays flat or lower fees and older stock pays more, so the slow-mover check above matters more after that date. The "long-term storage fee" named in Amazon's Seller Assistant example is the older name for the aged inventory surcharge, which Amazon's fee pages now list under that name.

Amazon also publishes the reports an AI or a script can read, so you need not copy numbers by hand. Its Selling Partner API documentation lists a Stranded Inventory report that includes recommended actions, a Reserved Inventory report that splits reserved units into customer orders, warehouse processing and staging, a Storage Fees report with the estimated monthly storage fee for each ASIN, and an Inventory Ledger detailed report that covers 18 months of movements such as receipts, returns, removals, damage, loss and found units. The ledger is the one to use when a count is short, because it shows where the units went.

Each task is a narrow read of data you already have. That makes it a safe use of AI. Asking a model to decide what to restock is riskier, because it does not know about your supplier delays or a launch you are planning.

A real example: Amazon's own Seller Assistant

Amazon has built this idea into Seller Central, the dashboard where sellers manage their Amazon business. In September 2025 Amazon announced an upgraded Seller Assistant, an AI helper that can reason, plan and, with the seller's permission, take actions. Amazon says it is powered by Amazon Bedrock (Amazon's cloud service for AI models), using Amazon Nova and Anthropic Claude models, and that it is free to sellers.

On inventory, Amazon says Seller Assistant will actively monitor inventory levels and alert sellers to actions they can take. Its own example: while a seller reviews FBA inventory, it flags slow-moving products before they incur long-term storage fees, and recommends which to leave alone, mark down or remove. It is also meant to analyze demand patterns and prepare shipment recommendations. Amazon described several capabilities as rolling out over the months after the announcement, so check what is switched on in your own account.

Note what this example shows and what it does not. It shows the big platform treats inventory monitoring as a good job for AI. It does not show a published result such as a percentage reduction in stranded stock. It also only sees Amazon's side. If you sell on other channels too, your full picture still lives in your own records.

How to do this yourself with Claude

Claude is a general AI assistant made by Anthropic. This works with the free version, though a paid plan handles bigger files.

1. Download the report. From Seller Central, open the FBA Inventory page or Manage Inventory and export your inventory as a spreadsheet or CSV file. Remove anything you do not need, such as buyer details.
2. Paste or upload it. Give Claude the file.
3. Ask for exceptions. Type: "This is my Amazon FBA inventory report. List every SKU that is stranded, unfulfillable, or where available plus reserved plus inbound does not match the total. Add a column explaining why each is flagged. Do not guess about anything the report does not show."
4. Ask for the aging risk. Type: "Using units on hand and units sold in the last 30 days, list SKUs with more than 90 days of supply, sorted by units on hand."
5. Draft the case. For one flagged SKU, type: "Write a short Seller Support case for this stuck inventory. Include the SKU, ASIN, units and what I am asking Amazon to do. My notes follow after this sentence."
6. Check before acting. Open Seller Central and confirm each flag. The model may misread a column or a status name.

A worked example with made-up numbers: your report has 3 SKUs. SKU A shows 40 available, 5 reserved and 10 inbound, total 55, which adds up. SKU B shows 30 available, 0 reserved, 0 inbound, total 38. The model flags B because 8 units are unaccounted for, which you then check in Seller Central. SKU C shows 60 stranded units and 0 sales in 30 days. The model flags C as the most urgent because it is costing storage fees and earning nothing.

For stranded stock, Amazon's Fix Stranded Inventory tool is the first place to look. An older Seller Central announcement describes automatic actions there: automatic relist for reasons such as a closed listing, and automatic change to FBA for merchant-fulfilled listings, with a default wait of 30 days that you can shorten or opt out of. That post is several years old, so open Fix Stranded Inventory in your account to see the current settings. If you want a plan for where else AI can help in your selling business, the AI Strategy Sprint is a paid 3-day first step, starting with a free 15-minute AI brainstorm call.

Where a person still decides

• Whether to relist, remove or mark down a stuck SKU.
• Whether a count gap is Amazon's error or your own receiving mistake.
• How much to reorder. That depends on supplier lead times and plans the model cannot see.

How I'd approach this

Start with the free weekly check. Pick one day, export the report, run the prompts, and fix the top three flags. After a month you will know which problems repeat. If the same ones keep coming back, or you sell on several channels and keep reconciling by hand, that is when a small connected system earns its keep: one that pulls Amazon and your other channels into a single stock list and sends an alert when something drifts. It does not need to be big. See tool stack for an ecommerce store for how the pieces fit, and what an AI agent is if you are curious about the automatic version.