The client
A waste management and industrial services company serving hundreds of B2B clients — construction sites, facilities, industrial plants — with regular supply needs (containers, equipment, consumables).
The problem
Clients were placing orders via WhatsApp in every format imaginable: photos of handwritten lists, voice notes, free-text messages, sometimes just “the usual for next Tuesday.” A human operator had to read each message, interpret the order, look up product codes, and manually enter everything into the ERP.
The process was slow (15–20 min per order during peak times), error-prone, and didn’t scale. Clients couldn’t order outside business hours.
What we built
A WhatsApp-to-ERP bridge powered by AI:
- Clients send orders via WhatsApp in any format — text, photo, voice note
- An AI agent interprets the message, extracts product names and quantities
- Agent validates items against the product catalog, flags ambiguities (e.g., “small container” → asks which size)
- Confirmed orders push directly into the ERP via API
- Client receives a WhatsApp confirmation with order summary and expected delivery
Tech stack: WhatsApp Business API, Claude API, custom catalog matching engine, ERP REST integration.
The result
- 95% of orders processed with zero manual entry
- Order processing time: from 15–20 min to under 90 seconds
- 24/7 ordering — clients no longer wait for office hours
- Error rate reduced by 78% compared to manual entry (measured by corrected/canceled orders)
- Operator workload shifted from data entry to customer relationship work
Order volume grew 31% in the 6 months following deployment, without adding headcount in the order management team.