One wrong box costs more than the whole order.
Catching mistakes at the bench and finding out why the same ones keep happening.
This is the delivery note for today's inbound load. Turn it into a table with item code, quantity and pallet number.
Claudereads the lines off the photo
Codexchecks codes and counts against the order
Geminiflags lines that could be misread
Inbound load, 2 Oct
- 42 lines read into a table: code, quantity, pallet
- 2 lines unclear in the photo, marked for you to check
- Pallet 7 has 3 items fewer than the order
The order said twelve. The box had ten. Nobody noticed until the customer did, and now there is a return to pay for, a replacement to ship and an apology to write.
- Returns, reshipping and penalties on thin margins
- Hours spent typing delivery notes into a system
- The same error repeats because nobody counts them
Try it yourself in Luminair
Six steps. No setup beyond the AI plan you already have.
- 1
Install Luminair and sign in
Download Luminair and open it. Go to Settings → Models → Add account and sign in with an AI plan you already pay for, like Claude, ChatGPT or Gemini.
- 2
Make a folder for this job
In the sidebar, choose New folder… and name it “Receiving”. Then choose New session in this folder. Everything for this job stays together.
- 3
Bring the information in
You need: photo of the delivery note and the purchase order. Connect Gmail, Airtable and Trello in Settings → Connectors, or export the files and drag them into the chat.
- 4
Ask for the result
Paste this into the session, or say it in your own words:
This is the delivery note for today's inbound load. Turn it into a table with item code, quantity and pallet number.
- 5
Get a second opinion
Pick a different model in the model picker and ask it to check the first answer:
Which lines in this table could have been misread from the photo?
- 6
Make it repeat
Open Flows → New and add the steps: Read the photographed delivery note → Build the table and flag unreadable lines → Compare with the purchase order. Choose when it runs, like every Monday at 09:00. Nothing is sent or changed until you approve it.
How it plays out
Photograph the paperwork
Take a photo of the delivery note on your phone and send it to the Receiving folder. Ask for the lines as a clean table: item code, quantity, pallet. Check it, then paste or import it into your system.
Check the tricky orders
For orders that go wrong most often, photograph the packing list next to the box. Ask whether counts and labels match the order. A mismatch is flagged before the tape goes on.
Count the mistakes
Log each error in Journal in one line. Once a month, ask what they have in common: a product, a shift, a client whose instructions are unclear.
You stay in charge
A photo check helps, it does not replace a person counting. Luminair does not plug into your warehouse system directly, so you move the checked table across yourself.
The fix is usually small: a clearer label, a different bin, a second look on one client's orders. You find it because the errors are finally written down in one place.
Sign in with the AI plan you already have and start with this job.