Monsoon is Fairshift's trade-intelligence and orchestration layer. It carries approved product knowledge and commercial context across buyer conversations, classifications, quotations, trade documents, certificates, and booth transcripts. Version numbers are working labels, not claims about a proprietary foundation model.
The build line so far
- 011.0, the apprentice: established the paperwork context. HS classifications, quotation formats, and the documents that move goods across borders.
- 022.x, the linguist: carried approved knowledge into more buyer languages without losing the commercial record.
- 033.x, the operator: routing matured across specialist workflows, configured reasoning, and human review.
- 043.3, current: coordinates context across web, WhatsApp, voice, email, kiosk and capture.
What the current build is better at
Three things, honestly stated. Classification confidence is expressed rather than implied: when the workflow is unsure about an 8-digit split, it shows the candidates. Quotation drafting holds your house format more faithfully across languages. And routing between trade logic, configured reasoning, and human review is more explicit.
Its limits also matter. Monsoon is not a general assistant and does not give final legal answers. It will not invent a buyer. If a trade listing has no proof of buyer volume, Monsoon says so.
The Monsoon page covers its working context, routing, approval boundaries, and a day on the desk.
Inside Monsoon