Inventory Reorder
It's Monday, and the spreadsheet that keeps you in stock is one person's job. This agent reads on-hand against your reorder points, subtracts what's already on order, checks it at the variant level, and rounds up to case pack. Then it drafts the replenishment order — vendor, SKUs, quantities, total — and stops there. It never sends it. A human reads the draft and decides.
Yours outright — no required subscription.
Every agent is built for your business — your systems, your approval chain, your way of working. We scope it on the call.
If this sounds too technical, don't worry. We take care of everything for you.
- Private VPS deployment. This agent runs on a server that belongs to you, not a shared cloud tenant. Your cost basis, vendor terms, and sales history stay on your box.
- A technician agent alongside it. Every hire ships as a pair: the Inventory Reorder Agent watching stock and drafting orders, and a VPS-technician agent on the same box keeping it healthy — patched, backed up, monitored. You're not hiring one thing, you're hiring a small team of two.
- Connection to where your stock numbers already live. Wired into your inventory system and the sheet you already keep reorder points, vendors, case packs, and lead times in, so nothing has to be re-entered somewhere new. Most of setup is getting open-PO quantities into a field it can actually read.
- A draft order per vendor. Flagged SKUs grouped by supplier, quantities rounded to case pack, with the running total in front of you so you can see where an order lands against a freight minimum or a price break before you send it. The draft waits; it does not go out on its own.
- Security hardening. Private VPN, firewall, and encryption configured on your server before it reads a single stock number.
- A live walkthrough, then 14 days of priority support. We sit down on a call and show you what each agent does and where the reorder agent is supposed to stop. Then two weeks of priority support to tune the velocity window and the per-vendor rounding rules against a real month of orders.
- You own everything — the server, the agent, the code it runs on.
Not sure it fits? Check fit in 90 seconds in the free assessment chat.
Who this is for
- You're a small brand with a few hundred SKUs across variants, and one person exports the inventory list every Monday, pastes it into the same spreadsheet, and eyeballs the bottom of the sorted column. When they're out that week, nobody does it.
- You can't say what's already on the water. Open purchase orders live in email threads and a folder of PDFs, so answering "did we already order this?" takes three tabs and a phone call — or you skip it and order the thing twice.
- You sell the same physical units through your store, a marketplace, and a few wholesale accounts, and one 400-unit wholesale line can drain six weeks of retail stock in an afternoon without anyone noticing until it's gone.
- You run a practice or a trade shop where someone walks the supply closet with a clipboard, or runs a par sheet against every truck before the supply-house run, and every SKU missed on that sheet becomes a tech idling in a parking lot at 10am.
How it earns trust
The failure mode of AI in replenishment isn't that it gets a number wrong — a spreadsheet does that too. It's that you can't tell it was wrong, or which input caused it, until the shelf is empty. This agent is built on the opposite bet: every flag it raises shows the arithmetic underneath it.
Every flag is a stamped row
When it says reorder SKU-4471, the record shows the on-hand reading and when it was read, the reorder point and where that number came from, the velocity window in units per day and over which dates, the lead time assumed, any open-PO quantity subtracted, and the rounding that turned 150 into a case-pack 168. You can argue with it line by line, because you can see which input was wrong.
Silence is logged too
Suppressions are recorded the same way flags are: "did not flag SKU-991, excluded by Dana on 6/12." An item going quiet is visible instead of invisible, which is how most reorder spreadsheets fail without anyone noticing for a quarter.
The stockout post-mortem
The real test is the day you run out anyway. You can look up whether it flagged, when, and what happened next. Either it never flagged and the log names the bad input — usually an on-hand number or an open PO it couldn't see — or it flagged on the 3rd and the draft sat unapproved until the 12th. That second answer is a process problem, and it's the one nobody can get out of a spreadsheet today.
None of that makes it right more often; it makes it checkable, which is the part you can act on.
Pairs well with
These share a workflow with this role. Tick any to add them to your setup.
The system says we have 40 and there's nothing on the shelf. What good is this?
It reorders against a number, not a shelf, and it's only as right as your count. Damages nobody recorded, a return never restocked, a case opened for a photoshoot, plain shrink — any of those make it confidently wrong in the direction that hurts, because it stays quiet while you run out. It also never edits or reconciles a quantity on its own; the fix for a bad number is a human recount. What it does give you is the on-hand reading it used and the timestamp it read it, so when the shelf disagrees you can see exactly which number lied.
I don't actually know what's on order right now. Can it?
Only if the open quantity lives somewhere it can read. If a PO exists solely as an email thread with your rep, the agent has no idea 400 units are on the water, and it will re-flag the same SKU every morning until someone orders it twice. Getting open quantities into a field — a column in the sheet, a status in your system — is usually the real setup work, not the math. We scope that on the call before anything gets built.
Does it ever place the order by itself?
No. Every replenishment ends as a draft you read and send. It cannot commit your money to a vendor, it does not log into a supplier portal, and it does not email your rep, chase a late PO, or talk to your customers. It also never touches a product listing to hide an item or post a restock date. Reorder points optimize for not running out; they know nothing about a slow month or a tax bill, so the cash decision stays with the person who has to make it.
The reorder points were set two years ago by someone who doesn't work here.
Then they're wrong, and the agent won't quietly fix them — it doesn't set or change reorder points, vendor assignments, or exclusions on its own. Those are your numbers in your sheet. What it does is show its work against them and keep its suggestion next to what you actually ordered, so after a month you can see where it runs high or low and retune the inputs instead of trusting or distrusting the whole thing. Marking a SKU excluded stops the flag, and the suppression gets logged so an item going quiet is visible rather than silent.
We're either out of the stuff that sells or sitting on the stuff that doesn't. Does averaging fix that?
Not by itself. A thirty-day velocity window is a fine input until a viral week, a promo, or one big wholesale order sits inside it — feed it a spike and it reorders for demand that isn't coming back. It can't tell a trend from an event unless you tell it which weeks to distrust, so the window and the exclusions get set with you during setup and adjusted in the first two weeks. Vendor reality is the same story: MOQs, case packs, a price break at 500, allocation. Those rounding rules get entered per vendor by hand, because nobody publishes them.
It's all in Kevin's head, and Kevin's on vacation. What's covered under the $1,997 setup?
One company, one system of record, one approval chain — flat, one time. The point is that the Monday loop runs the same week Kevin is gone, with the reasoning written down instead of remembered. Scope is set on a free 15-minute call, and you're typically live about seven days from kickoff. You own the server, the code, and the keys; we keep the recipe we used to cook it. Multiple warehouses with genuinely different replenishment rules is a scoped conversation, not a price adjustment.
Hiring more than one? A department on tap — the subscription puts a build team behind every request, agent after agent.