The problem
A request for quotation asks suppliers to price technical items. Staff were pasting prompts into a chatbot, copying the answers, and fixing them by hand. This took up to 30 minutes per group and created a large backlog. The new system had to cost less than doing it manually. Quality came first, though. Every item had to be realistic and complete enough to buy without the vendor needing to ask questions.
Our approach
We started with a paid manual test, built a spreadsheet version, then one owned web app. Our design rule: AI decides what to ask for, and plain code enforces how. We advised against a separate keyword tool because the main pipeline already handles vague input.
What we built
We built a web app that takes keywords and writes all five enquiries in under 90 seconds. It checks the work and saves a finished document to the client's shared drive. People still source the keywords, review the final output, and send the documents from their own email.
The results
Before
10 to 30 minutes of copying, pasting, and fixing a chatbot's answer by hand.
After
30 to 90 seconds to get a checked RFQ document in the client's template.
- 30 to 90 seconds, down from 10 to 30 minutes.
- Every group passes an AI check and 18 code checks before review.
- All 50 varied groups passed our internal test.
All figures come from our own project records and are not independently audited.
Next storyOne product record in, listings for every marketplace out