MSJ is a leading ship supply company specializing in vessel restocking services when ships dock at port. When the company identified an opportunity to dramatically accelerate their RFP response process, they turned to Dreamix to develop an AI-powered solution. The result? MSJ reduced their RFP processing time from 1.5-2 hours down to just 3-5 minutes, while maintaining accuracy levels comparable to their experienced staff.
The Story of our Partner
MSJ is a leading US-based ship supplier, providing vessels with everything from food and beverages to consumables and essential supplies. Part of their business model is responding to Requests for Proposals (RFPs) from ships that need restocking while in port.
With over 9,000 products in their warehouse management system, MSJ has built an extensive inventory to meet the diverse needs of various vessels.
The Challenge
In the maritime industry both speed and accuracy matters. MSJ recognised that usually, the first supplier to respond to an RFP is the one to win the contract.
Every time a ship submitted an RFP, MSJ needed to match the requested items against their warehouse inventory system to verify availability and pricing. However, product names and units of measurements vary between the RFP and their internal system. This meant MSJ employees had to manually open each RFP alongside their warehouse system, compare items line by line, and create a detailed proposal with pricing information for every match. The entire process took 1.5-2 hours per RFP.
In an industry where the fastest and most accurate responder wins the business, MSJ recognised that this was costing them potential business. They needed a way to process RFPs faster without sacrificing accuracy.
The Dreamix Solution
Dreamix began by designing a prototype to validate requirements before development. The solution needed to handle unique Excel template formats that MSJ regularly encountered. The team took an incremental approach - we perfected the complete workflow and expanded it to all templates.
Initially, we used traditional string matching algorithms, removing special characters, normalising text and weighing word significance. That only achieved 30% accuracy in tests, which wasn’t enough for us.
So, we pivoted to AI. The team deployed a machine learning model that creates vector embeddings for all 9,000+ warehouse products. When an RFP is uploaded, the system vectorizes those products too, then matches items by finding the closest vectors. This semantic approach understands meaning rather than just matching text, allowing it to recognize that "orange juice" and "100% orange extract with pulp" are the same product.
The final solution reads uploaded RFPs, performs AI-powered semantic matching against warehouse inventory, extracts pricing, and generates complete proposals ready for export. A human-in-the-loop feature lets MSJ employees review and correct matches before confirming. Every correction trains the system, making it smarter over time.
The Results
Thanks to our work together, RFP processing time for MSJ dropped from ~1- 2 hours to 3-5 minutes - a time reduction of over 95%. This allows MSJ to react to RFPs much faster than competitors, directly translating to more won contracts.
The AI system is currently over 95% accurate, and perfecting itself further with every iteration and correction made by staff.
By automating the tedious manual work of product matching, MSJ's team can now focus on higher-value strategic activities.
