A raw material deal usually starts as a message. An enquiry usually arrives by email, and before anything can happen with it, someone retypes it into a system. The offer that responds to it goes back out as a document that no system ever sees again. Most of the work in between is transcription, and none of it is the commercial judgement anyone is paid for.

The two ends of that path are not the same problem. Reading an incoming enquiry and turning it into a structured record is something a model does well today. Drafting the offer that responds to it, with the commercial judgement inside it, is a harder problem and a later one.

This episode follows one deal through both ends of that path, from what the capture step can do today to where the human judgement still has to sit.

MetalshubTalks 010 showed what becomes possible when a company points its own AI at its Metalshub data through the MCP connector, which needed an AI licence. This session starts from the other end, inside the platform.

The discussion explores:

• Where the manual work actually sits between an incoming enquiry and an outgoing offer
• What an AI can pull out of a plain-text message, and what a person still has to check
• Why the capture step decides whether anything downstream can be trusted
• What changes commercially once an off-platform deal has a structured record behind it
• What Metalshub is building next
• A live walkthrough of the flow, from an incoming enquiry to the offer that responds to it

MetalshubTalks host Davide Nava is joined by Dr Sebastian Kreft, Co-Founder and Managing Director of Metalshub, and Julien Dedek, Head of Platform AI at Metalshub.

This session is designed for procurement leaders, commercial and sales teams, traders, and supply chain managers who want a practical view of what AI changes in day-to-day raw material trading.