Email to shipping label. Webhook (email text) → AI extract → Design → Print.
AI / WORKFLOW / OCR
When the Data Is Messy, Let AI Fill the Label
Turn free text, order emails or a photo of a supplier label into clean label fields. LabelInn's AI workflow step extracts, classifies and reads images, and fails loudly instead of guessing.
FREE PLAN · NO CREDIT CARD · WINDOWS / MAC / LINUX / WEB
THE PROBLEM
Not every label starts from a clean spreadsheet. The order arrives as a paragraph in an email. The supplier's batch number is on a photo of their label. The product description says "corrosive, keep away from children" in free text and someone has to decide which hazard label applies. That reading and deciding is where time goes.
The AI node in a LabelInn workflow takes the record, follows your instruction and returns structured fields that the label can use.
🤖 Messy in, label-ready out. And if the AI is not sure, the workflow stops instead of printing a guess. Start free →
01 / AI
Four modes
| Mode | What it does | Example |
|---|---|---|
| Extract from text | Pulls named fields out of free text | Customer, address and quantity from an order email |
| Read an image | Reads text from an image URL or upload | Lot and expiry from a photo of a supplier label |
| Classify | Picks exactly one of the labels you list | Which hazard category, which label size |
| Transform | Follows an instruction and returns JSON fields | Shorten a product name to fit 30 characters |
02 / EXTRACTION
Built so it cannot quietly print nonsense
- Classification is checked. The answer must be one of the options you listed. Anything else fails the step instead of being routed somewhere wrong.
- Bad answers fail. An empty reply, a reply that is not valid JSON, or a model error fails the node, and the workflow's error branch can catch it.
- Test runs are free. Testing a workflow simulates the AI step, so you can wire and try it without spending credits.
03 / OCR
Typical workflows
Goods-in from a photo. Photo of the supplier label → AI read image → Lookup in your product list → Design → Print.
Pick the label by meaning. AI classify the description → Router → the matching design. See conditional labels.
04 / CLASSIFICATION
Cost
Each AI step uses your company's monthly AI credits, one charge per call. Every plan includes a monthly allowance; see pricing.
QUESTIONS / FAQ
Frequently asked questions
Can AI extract label data from an email or free text?
Yes. The AI node's extract mode pulls named fields such as customer, address or quantity out of free text, and the label is filled from them.
Can it read a photo of a label?
Yes. The image mode reads text from an image URL or upload, for example a lot number and expiry date from a supplier's label.
What if the AI gets it wrong?
Classification answers must be one of the options you listed, and empty or malformed replies fail the step. The workflow's error branch catches the failure instead of printing a guess.
Does it cost extra?
Each call uses your monthly AI credits. Test runs of a workflow simulate the AI step and are free.
KEEP GOING
Related
NEXT STEP / LABELINN
Print from data that isn't tidy yet
Add one AI step between the messy input and the label. · Extract, read images, classify, transform · Fails loudly, never guesses · Free test runs