The challenge
Nam Leong supplies pipes, fittings, and valves across six industries, from fire protection to oil and gas. Customers send requests for quotation in every format, from Excel and PDF to photos of handwritten lists, and each line has to be matched by hand against a catalogue of more than 100,000 SKUs, each with its own abbreviations and quoting rules. Preparing a single standard quote took hours, and most customer questions arrived over WhatsApp.
What we did
- Built an AI quotation platform on Google Cloud with a four-agent Gemini engine: one agent reads any file format with multimodal vision, one matches each line against the catalogue with retrieval-augmented search, one recommends add-on services such as cutting and threading, and one applies pricing rules
- Digitised supplier catalogues automatically with Document AI, and vectorised the full SKU master list with Vertex AI embeddings into a pgvector database for hybrid semantic and fuzzy search
- Delivered a human-in-the-loop dashboard so sales staff review the AI's matches and confidence scores before a quote goes out, with branded Excel and PDF export
- Added a WhatsApp Business chatbot, a Gemini agent with function calling into the same product database, with fact-checking against live data and human handoff for complex questions
The results
- A quotation platform that ingests any RFQ format and returns matched, priced quotes with a clear audit trail
- A WhatsApp chatbot that answers product, specification, and availability questions from the live catalogue
- Infrastructure delivered as code, with the AI grounded on Nam Leong's own data so answers stay accurate