Golden International Corporation
Food importer and distributor, Renton, Washington
- Size
- Not stated
- Where
- US
The job
A consultant built an order entry agent for the firm using Gemini and n8n. It takes orders sent as handwriting, photos, PDFs or typed documents, reads them, matches each line against the product catalogue of over a thousand items, scores its confidence and sends high confidence orders straight into the old ERP system, flagging the rest for a person.
Tool: Custom Built (Gemini and n8n) · More using Google Gemini
What happened
Order entry fell from up to two hours per order to about five minutes.
Typed orders are read at nearly 100% accuracy, order processing capacity more than doubled, and the agent is projected to cost about $20 a month at current volumes.
Before you believe it
Prowess Consulting built the system and published the story, so the times and accuracy figures are the builder account. Accuracy is about 90% on handwritten orders, so the human check on low confidence lines is part of the result, not an optional extra.
Try this prompt
Test the job on your own material before buying anything. Replace the highlighted brackets.
Here is an order a customer sent us: [ORDER]. Read it, match each line to an item in this catalogue: [CATALOGUE]. Give every line a confidence score. Draft the ERP entry for lines you are sure of and list the lines a person must check, with the two closest catalogue matches for each.
Source
Prowess Consulting ↗, 1 June 2026 · vendor story
This source was written or published by the supplier of the tool, so treat the claim as a sales story.
More from the archive
Mothership Coffee Roasters
Coffee roaster with six cafes, Las Vegas
The founder connected the till system, QuickBooks, Slack and email to Claude and made it the leadership team workspace for forecasting and weekly planning across six cafes plus wholesale and catering.
Result In-store margins rose to 22%.
Not statedUS
Russell Roof Tiles
Roof tile manufacturer, Burton on Trent and Lochmaben
The firm built its own internal AI tools over the past year. Operators record quality checks digitally from the factory floor with real-time performance data across two sites, staff query the ERP system directly to find priority orders, and month-end reporting tasks run automatically.
Result One month-end reporting task fell from 60 to 90 minutes to about three minutes.
About 160 staffUK
Reno Salvage
Salvage and scrap metal yard, Reno, Nevada
The operations manager uses ChatGPT to troubleshoot equipment faults as they happen, to look up specifications and materials for customer questions on the spot, and to build a part numbering system for the yard stock.
Result A part numbering system for over 1,000 items was built in a single afternoon.
Not statedUS
Keep browsing: more in Operations (14) · more using Google Gemini (4)
New entries every week
The archive grows weekly. Get the stories in your inbox, with the context behind them.