Background

Ampwise brokers industrial goods between buyers and suppliers in Europe.

Challenge

Comparing quotes from emails by hand took up most of the staff’s day.

Solution

We built a system that reads emails, interprets supplier price quotes in any format and finds suppliers automatically. Development took six months.

  • PostgreSQL
  • Docker

Result

According to Ampwise, while it was using the solution its quarterly turnover grew from €70,000 to €1.8 million in two years, and productivity increased tenfold.

Lesson learned

This was our first large project with language models. The main challenge was stopping the AI from making things up; the answer was more precise instructions and feedback loops in which users can spot and correct errors.

Book an initial call (opens in a new tab)All work