An AI inventory manager is software hired to keep track of your stock between counts. It reads your inventory records, keeps the count current as parts go out on jobs and deliveries come in, notices which lines are running low and which have not moved in months, and puts a reorder in front of a person early enough to act on. It recommends. A person orders. When a count does not match what is on the shelf, it tells someone instead of fixing the number quietly.
Most of what gets sold as AI for inventory management is demand forecasting for large retailers: models trained on years of sales history across thousands of stores. This post is about something smaller and more common. You have a stockroom, a van, or a warehouse shelf, a spreadsheet or an inventory app that is roughly right, and a habit of finding out you are out of something on the morning you need it.
At Sanaf this is the AI inventory manager job, one of the jobs on the roster a business can give Sanaf. Where the job description says something plainly, this post quotes it.
Why stock runs out at the wrong moment
Ask anyone who has run a stockroom what goes wrong and you rarely hear "we could not count." You hear that a common part ran out halfway through a job, or that somebody over-ordered after the last time that happened and now there are forty of something on a shelf that nobody has touched since spring.
Those are the same failure seen from two ends. In both cases the business found out at the wrong time. The job page puts it in one line: "Counting stock is not the job; knowing early enough to reorder before a job stops is the job, and that is a timing problem rather than a counting one."
It also names the signs you have this problem:
- Jobs get delayed because a common part ran out.
- Stock counts happen once a year and are wrong by February.
- Cash is sitting on shelves in things that never move.
If you recognise one of those, the rest of this post is about what an AI inventory manager does about it.
What an AI inventory manager does in a normal week
The job has six standing duties. Any one of them is something a careful person already does when they find a free afternoon. The difference is that they get done every day.
It tracks inventory against your records
It works from the inventory records you already keep, whether that is a Google Sheet, Zoho Inventory, or the inventory side of QuickBooks. It does not need barcodes or a new system to start. The job page is direct about this: "It works from the inventory records you already keep, and for a lot of businesses that is a spreadsheet and a supplier list."
It monitors stock levels
As job records show parts going out and deliveries show parts coming in, the count moves with them, so the number you look at on a Tuesday includes Monday's jobs.
It identifies low inventory while there is still time
Low is relative. Ten of a part you use once a month is plenty. Ten of a part you use four of a day, from a supplier who takes a week and a half to deliver, is a problem already. So the question it asks about each line is whether the stock will last until a new order could arrive, and it flags the ones that will not.
It recommends reorders with the working shown
When a line needs ordering, the recommendation arrives with the quantity, the supplier and the reason already filled in: how fast it has been moving, how long the supplier usually takes, and how much is left. Approving it should take a moment, because the thinking is already on the page.
It tracks inventory movement, including what does not move
Watching what does not move is as much of the job as watching what does. Money tied up in stock that has sat for a year was needed somewhere else, and nobody notices because nothing goes wrong. The job keeps that list in view.
It generates inventory reports
It produces the report nobody has time to build by hand: what is low, what was reordered, what has not moved, and where the count and the shelf disagree.
Sanaf does all of this inside Slack or Microsoft Teams, where your team already talks, and reaches only the apps you connect for the job: your inventory records, your supplier list, your job records and your accounting software, plus Gmail for supplier email.
How AI in inventory management decides when to reorder
It helps to see the arithmetic, because it is the same arithmetic a good inventory manager does in their head.
The reorder point formula
The reorder point is the stock level at which you should place an order so the delivery arrives before you run out. The standard reorder point formula is:
Reorder point = (average daily usage x lead time in days) + safety stock
Say you use 4 of a particular fitting a day on average, and your supplier takes 10 days to deliver. Lead-time demand is 40. If you reorder at 40, you run out on the day the delivery is due, which only works if every day is average and every delivery is on time.
Safety stock
Safety stock is the cushion for the days that are not average. One simple safety stock formula is:
Safety stock = (maximum daily usage x maximum lead time) - (average daily usage x average lead time)
If your busiest days use 6 of that fitting and your supplier has taken as long as 14 days, that is 84 minus 40, so a safety stock of 44. Your reorder point becomes 40 plus 44, which is 84.
There are more statistical versions of the formula that use the standard deviation of demand and a target service level. They are better when you have clean history. Most businesses do not, and the simple version is honest about that.
Where AI helps
The formula is easy. Keeping the inputs current is the hard part. Average daily usage drifts with the season. A supplier who used to take 10 days now takes 16. A new kind of job starts using a part that used to be slow. An AI inventory manager recalculates from the actual job records and delivery history instead of a number someone typed into a spreadsheet two years ago. That is what "AI demand forecasting" means at this scale: noticing that the inputs moved, and moving the reorder point with them.
Dead stock, excess inventory and cash on the shelf
Dead stock is inventory that has not sold or been used in a long time. There is no universal cut-off; many businesses use six or twelve months. Excess inventory is the broader category: more of something than you will use in a reasonable period, even if it does move a little.
Both are expensive in a way that never shows up as a line on a report. The cash is spent, the shelf space is taken, and the parts slowly go out of date or get damaged. Over-ordering is the natural reaction to running out, which is why the two problems travel together.
An AI inventory manager handles this side by keeping a standing list of slow-moving inventory, with the date each line last moved and what it is worth at cost. What to do about it is a business decision. You might return it to the supplier, use it up on the next suitable job, sell it off, or write it down with your accountant. The job's part is making sure somebody sees the list.
This also connects to inventory turnover, which is cost of goods sold divided by average inventory for the period. A low turnover number usually means too much money is sitting still. You do not need to track the ratio to act on it; the list of lines that have not moved tells you the same thing in a form you can do something about.
Cycle counting instead of one annual count
Most businesses do a full physical inventory once a year. By February, as the signal above says, the number is wrong. Cycle counting is the alternative: count a small slice of stock on a rotating schedule, so every line gets counted several times a year without closing the doors for a day.
The usual way to decide what gets counted most is ABC analysis. The few lines that make up most of the value, or that stop a job when they run out, get counted often. The long tail of cheap, slow items gets counted less.
The comparison, cycle count vs physical inventory, comes down to this:
| Annual physical inventory | Cycle counting | |
|---|---|---|
| How often each line is counted | Once a year | Several times a year, more for important lines |
| Disruption | A day or more of counting | A few minutes a day |
| When errors are found | Up to a year later | Within weeks |
| How accurate the count is in February | Drifting | Close |
| What it needs | A free day and every hand available | A schedule and someone who follows it |
The schedule is the part that slips. An AI inventory manager can keep the cycle count rota, post the day's short list of lines to count in the team channel, and compare the results against the records. It does not walk the shelves. A person does, and the job makes sure the counting happens and the results go somewhere.
Shrinkage, mismatches and the rule it works under
When the count and the shelf disagree, something happened. Maybe a part went out on a job without being recorded. Maybe a delivery was short. Maybe it was damaged, or miscounted last time, or stolen. That gap is called inventory shrinkage, and it is information about the business.
This is where the job's hands-off rule matters. It is short enough to quote in full: "You recommend, a person orders. Anything that looks like shrinkage or a count that does not reconcile is flagged to a person rather than quietly corrected."
The second sentence is the unusual one. A lot of automated inventory tools adjust the count to match the shelf and move on, which looks tidy. The job page explains why Sanaf does not: "A discrepancy is information about the business, and an automated count that quietly corrects itself destroys that information while looking tidy." So a mismatch goes to a person with the history that led to it: when the line was last counted, what went out since, what came in.
The first sentence covers money. Placing an order commits cash, so it is something the job proposes and a person approves. Sanaf's rule across every job is to ask before anything it cannot undo and to keep a record of what it did. An order sent to a supplier is hard to take back.
What does an inventory manager do, and where does AI fit?
A human inventory manager in a larger operation does all of the above plus things software cannot: negotiating with suppliers, deciding what to stock at all, organising the physical layout, training people, and walking the floor. Plenty of businesses never had one. The job gets done by whoever runs operations, in between everything else, which is how a common part runs out.
An AI inventory manager covers the watching, the arithmetic, the reminders and the reports. It gives the person who does own stock a current count, a reorder list with the working shown, and a list of what is sitting still. Buying decisions and supplier relationships stay with that person.
It also pairs with the jobs next to it. Purchasing is a separate job from stock: comparing quotes and preparing orders is the procurement coordinator job, and a reorder recommendation can hand straight to it. The numbers side, including what stock is worth on the books, sits with bookkeeping.
An example week
Here is how it looks in practice for a trades business with a stockroom and three vans.
On Monday morning, Sanaf posts in the operations channel: two lines are below their reorder points. One is a pipe fitting used on almost every job, with 30 left, usage running at 5 a day this month, and a supplier averaging 9 days. The recommendation is 120 from the usual supplier, with the numbers beside it. The operations lead approves it in the thread, and the order goes to procurement.
On Tuesday, it posts the day's cycle count: eight lines, the fast movers first. Someone counts them in ten minutes. One line is short by six against the records. Sanaf flags it with the last count date and the jobs that drew from it since, and leaves the record as it is until a person looks.
On Wednesday, a supplier email says a delivery will be a week late. Sanaf recalculates, sees that one line will now run out before it arrives, and says so.
On Friday, the weekly report lands: what was reordered, what was counted, the one open mismatch, and a list of eleven lines that have not moved in over six months, worth a total at cost that somebody should look at.
Every step is in the record: what it noticed, what it recommended, what a person approved.
Frequently asked questions
What is an AI inventory manager? An AI inventory manager keeps your stock count current from your inventory and job records, flags low stock before it runs out, recommends reorders with the quantity, supplier and reasoning filled in, tracks slow-moving stock, and produces inventory reports. A person approves every order.
How is AI used in inventory management? At a large retailer, mostly for demand forecasting across many locations. For most businesses, AI in inventory management means keeping the reorder point inputs current, such as usage and supplier lead times, noticing when stock will not last until the next delivery, and keeping counts and reports on schedule.
Do I need barcodes or new inventory software? No. It works from the records you already keep, which for many businesses is a spreadsheet and a supplier list. A proper inventory system such as Zoho Inventory or QuickBooks makes it better, but noticing does not require a scanner.
Can it place orders with suppliers on its own? Not without a person approving. Ordering commits money, so the job proposes the reorder and a person decides.
What is the reorder point formula? Reorder point equals average daily usage times lead time in days, plus safety stock. If you use 4 a day, delivery takes 10 days and you keep 44 as safety stock, you reorder when you get down to 84.
What is cycle counting? Counting a small portion of your stock on a rotating schedule instead of everything once a year. Important lines get counted more often, and errors are caught within weeks.
What happens when the count is wrong? It is flagged to a person with the history behind it, and the record is not quietly corrected. A mismatch can mean shrinkage, an unrecorded job, or a short delivery, and somebody should know which.
How much does it cost? Sanaf starts free with $120 of credit and no card. After that you buy credits as you need them and pay for the work done, not per seat. The pricing page has the details.
The short version
An AI inventory manager works for the person who owns your stock. It gives them a count that is current, a reorder list that arrives early with the working shown, a cycle count that actually happens, and a list of cash sitting on the shelf. It orders nothing on its own and corrects no mismatch in secret.
Read the full AI inventory management job, see every job on the roster, or start free and give Sanaf your stock list.