I have been working in warehouse operations for a long time. Twenty-five years, give or take. In that time I have seen companies spend serious money on inventory systems that promised to solve everything. Most of them solved some things and created new problems. And very few of them made the people on the floor feel any less overwhelmed.
So when I started using Claude to help run cycle counts, I was not expecting much. I figured it would be a novelty. Something I tried once and moved on from.
I was wrong.
What I built is not fancy. It does not require IT. It does not need a six-figure software budget. It runs on a folder system, a spreadsheet, and a few well-written prompts. And it works. Every week.
Here is exactly how I do it.
The Problem With Most Cycle Count Systems
Cycle counts sound simple. Pick some parts. Count them. Compare the number to what the system says. Investigate the gaps. Done.
In practice, most teams skip them. Or they count the same easy-to-reach bins every time. Or the results sit in a spreadsheet that nobody looks at until something goes wrong.
The tools meant to fix this are often expensive, complicated, or both. Enterprise inventory management systems can cost more than a small warehouse earns in a month. Even mid-range warehouse management platforms carry steep implementation costs and training requirements.
Most small and mid-size operations do not have that kind of budget. And honestly, they should not have to.
The system I am going to walk you through costs almost nothing to run. You need Claude. You need Excel or a free spreadsheet tool. You need a folder on your computer. That is it.
What Is a Cycle Count and Why Does It Matter
A cycle count is a way to check your inventory in small batches on a regular schedule instead of shutting everything down for one giant physical count.
Think of it like this. Instead of cleaning your entire house once a year in a panic, you do a room each week. The house stays cleaner. You catch problems before they pile up.
Cycle counts do the same thing for inventory. You regularly count a portion of your stock. You compare what you counted to what the system thinks is there. You investigate discrepancies. You fix them. You move on.
Done well, cycle counts reduce shrinkage, improve accuracy, and give management something reliable to look at. Done poorly, they are just paperwork that nobody trusts.
This system helps you do them well.
Understanding the ABC Method (And Why It Changes How You Count)
Not all parts deserve the same attention. That is the core idea behind ABC classification.
It is based on the Pareto principle. In most warehouses, a small percentage of your SKUs account for the majority of your inventory value or movement. ABC classification sorts your parts into three groups based on that reality.
A Items
These are your high-value, high-priority parts. They make up roughly the top 10 to 20 percent of your SKUs by value or usage. They represent the most money. They get counted most often. In many operations, A items are counted monthly or even weekly.
B Items
These are your mid-range parts. Important, but not your top tier. They account for the next 30 percent or so of your SKUs by value. Count them quarterly or every other month. They matter, but they do not need the same frequency as your A items.
C Items
These are your low-value, slow-moving parts. They make up the largest portion of your SKU count but the smallest portion of your total inventory value. Count them once or twice a year. They still need to be in the loop, but they do not warrant constant attention.
Why does this matter for our system? When Claude pulls your 50-part count list each week, it uses your ABC classification to ensure the right parts are counted at the correct frequency. An item shows up more. C items rotate through less often. Your counting time goes where the money is.
What You Need Before You Start
This setup is simple. Here is what you need:
- Access to Claude (Claude Pro or Claude Cowork gives you the best folder integration)
- Your current inventory spreadsheet was exported from your ERP or CRM system
- A folder on your computer dedicated to this project
- Excel or Google Sheets for your count sheets and dashboard
- About an hour the first time you set it up
That is the full list. No special software. No plugins. No IT department required.
Step 1 โ Download Your Inventory Spreadsheet From Your ERP or CRM
Every cycle count starts with fresh data. That is not negotiable.
Before you do anything else, go into your ERP or CRM system and export your current inventory list. You want the most up-to-date snapshot of what the system thinks you have. If you count against stale data, your results will be unreliable before you even start.
The columns you need are straightforward:
- Part number or SKU
- Part description
- Location or bin
- Quantity on hand
- Unit of measure
- ABC classification (if your system tracks it)
- Last count date (if available)
Export it as a CSV or Excel file. Save it with the date in the file name. Something like inventory_export_2026_05_03.xlsx. That naming habit will save you headaches later when you are looking back at old counts.
Step 2 โ Set Up Your Folder and Drop In the Raw Data
Create a folder on your computer called something simple. Cycle Count System or Inventory Counts works fine. Inside that folder, create three subfolders:
- Raw Data โ This is where your fresh exports from the ERP go. Never edit files in this folder. Keep them clean.
- Count Sheets โ This is where your weekly count lists live after Claude generates them.
- Count Log โ This is where your completed counts and running history live.
Drop your fresh inventory export into the Raw Data folder. That is your starting point every time.
The reason you keep raw data separate is simple. You never want to accidentally overwrite or edit the source file. Raw data stays raw. Everything else is a working copy.
Step 3 โ Connect the Folder to Claude Cowork
This is where the system starts to feel different from anything you have used before.
Claude Cowork is a desktop tool that lets Claude read files directly from folders on your computer. Instead of copy-pasting data into a chat window, you point Claude at your folder, and it reads what is there. It can view your inventory spreadsheet, count logs, and history all at once.
Once your folder is connected, Claude has the context it needs to do real work. It knows what is in your inventory. It knows what has been counted. It knows what has not. You just have to ask the right questions.
If you are using the web version of Claude instead, you can upload your spreadsheet directly into a Project. Claude Projects work the same way. They keep your files and instructions in one place, so Claude has the full context every time you open it.
Step 4 โ Prompt Claude to Generate Your 50-Part Count List
This is the part that most people do not expect to work as well as it does.
You write a prompt. Claude reads your inventory file. It selects 50 parts using ABC logic. It gives you a count sheet ready to print or export to Excel.
Here is an example of the kind of prompt that works:
Using the inventory spreadsheet in the Raw Data folder, generate a cycle count list of 50 parts. Use ABC classification to weight the selection:25 A items, 15 B items, and 10 C items. Select randomly within each class. Do not include any parts that have been counted in the last 30 days based on the count log. Format the output as an Excel-ready table with columns for:Part Number, Description, Location, Expected Quantity, Count Quantity (blank), and Notes.
That is it. Claude reads the file, applies the logic, and gives you a clean list. You do not need to write formulas. You do not need to sort and filter manually. You ask. It delivers.
The random selection matters. When you let people pick what gets counted, they tend to count the easy things. The random pull inside each ABC class keeps the process honest.
Step 5 โ Print the List and Go, Count
Claude exports the count sheet as an Excel file. You print it. Your team goes and counts.
The sheet has everything the counter needs:
- Part number
- Description
- Location or bin number
- Expected quantity (what the system says is there)
- A blank column to write in the actual count
- A notes column for anything unusual
Keep it simple. The counter should not need any training to use the sheet. They look at the part number, find the bin, count what is there, and write it down. That is the whole job.
You can also go paperless. Some teams photograph the bin and log the count on a phone or tablet. Either way works. The key is getting the real number recorded.
Step 6 โ Enter the Count Data and Prompt Claude to Log It
When your team returns with the completed count sheet, you enter the actual counts into the spreadsheet. Then you bring that file back to Claude.
Here is the kind of prompt you use next:
I have completed the cycle count. The filled-in count sheet is attached. Please compare the actual count quantities to the expected quantities. Flag any discrepancies greater than 5 percent. Add all counted parts to the count log with today's date. List any flagged discrepancies separately so I can review them.
Claude reads the completed sheet, compares actuals to expected values, flags anything that is off, and updates your log. You get a clean discrepancy report without building a single formula.
The count log is the backbone of this system. Every part that gets counted gets recorded with a date. That is how Claude knows not to pull the same parts again next week. The log is what makes this a real system instead of a one-time event.
Step 7 โ Build an Excel Dashboard From Your Count Data
This is where the system starts to look impressive to people outside the warehouse.
Once you have a few weeks of count data in your log, you can ask Claude to help you build an Excel dashboard. You do not need to know how to build pivot tables. You describe what you want and Claude walks you through it, or generates the structure for you.
A basic dashboard for upper management might include:
- Total parts counted this week vs target
- Accuracy rate โ what percentage of counted parts matched the system
- Top discrepancies by value
- Parts not counted in 90 days (so nothing falls through the cracks)
- ABC coverage โ are you counting the right mix of A, B, and C items
You update the dashboard by dropping your latest count log into the folder and running a quick prompt asking Claude to refresh the data. Five minutes and you have a clean report ready for a Monday morning meeting.
Managers respond well to dashboards. It shows them the work is happening. It gives them numbers they can trust. And it positions you as someone who runs a tight operation.
Why This Works Even If You Have Never Used AI Before
You already know the hard part. You know inventory. You know what the numbers mean. You know what a good count looks like versus a sloppy one.
Claude handles the tedious part. Sorting through hundreds of SKUs. Applying classification logic. Comparing columns. Flagging outliers. Writing up a summary. These are things that used to take an hour and now take a minute.
You do not need to learn how to code. You do not need to understand how AI works under the hood. You need to know what you want and how to describe it clearly. That is the whole skill. And if you have been giving instructions to warehouse staff for any length of time, you already have it.
Think of Claude like a very capable assistant who can read your spreadsheet, apply your rules, and hand you back a finished work product. The assistant does not replace your judgment. It does the grunt work so your judgment has more room to breathe.
What This System Does Not Replace
I want to be straight with you here.
This system does not replace your ERP or WMS. Your NetSuite, your SAP, your Infor โ whatever you are running โ that is still where your inventory lives. This system works alongside it. You export from there. You compare counts against it. You push corrections back into it.
Claude is not reading your live system. It is reading a snapshot. That is why fresh exports matter. The system is only as good as the data you feed it.
This also does not replace the physical act of counting. Someone still has to walk the floor. Someone still has to look at the bin and write down the number. Claude cannot do that part. But it can make everything around that part faster and more reliable.
The Real Benefit โ You Skip the Expensive Software
Dedicated cycle count software exists. Some of it is good. Most of it is expensive. We are talking thousands of dollars a year for a mid-range solution. Enterprise-level platforms can run tens of thousands annually when you factor in implementation, licensing, and support.
This system runs on a Claude subscription and Excel. If you already have both, the cost is effectively zero, aside from the time to set it up.
For small and mid-size operations, that is the difference between doing cycle counts and not doing them. When a tool is too expensive or too complicated, it does not get used. When a tool is simple and cheap, it does.
That is the real win here. Not the technology. The consistency. Because a weekly cycle-count system is worth more than a perfect system that nobody uses.
Where to Start This Week
Do not overthink this. Here is your starting point.
This week, do one thing. Export your inventory list from your ERP. Any format works โ CSV, Excel, whatever your system can produce. Put it in a folder. Drop it into a Claude Project or connect the folder through Claude Cowork.
Then write this prompt:
I have uploaded my inventory spreadsheet. Can you review the data and suggest how I would set up an ABC classification if one is not already present? Then generate a sample cycle count list of 10 parts, weighted toward high-value items.
That is your first step. Ten parts. No commitment. Just see how it works with your actual data.
If it looks good, run it for real next week with 50 parts. The week after, enter your results and build the log. The week after that, pull the second count. Before the end of the month, you will have a working system.
It really is that simple.
Want Someone to Set This Up For You?
I know not everyone wants to build this from scratch. Some people would rather hand it off and have it ready to go.
If that is you, send me an email. I can put together a simple, custom setup based on your actual inventory data and your workflow. No bloated software. No long onboarding. Just a clean system that works the first time you use it.
Reach out at AI@ordinarypeople.com and tell me a little about your operation. I will take it from there.
Twenty-five years in warehouses taught me that the best systems are the ones that actually get used. This one does.
Ready to paste. The blockquotes (the lines starting with >) will render as styled quote blocks in WordPress. The headers, bullets, and bold text will all carry over clean. If your theme strips the horizontal rules (---), just delete those lines before pasting.
Joe Foley
Contributing writer at AI for Ordinary People, passionate about making technology accessible to everyone.
Leave a Reply