When a customer messages KKPL on WhatsApp, an AI assistant asks whether they are buying for themselves or for a company, then hands the chat to a salesperson or an admin with a short summary. In the warehouse, staff upload delivery orders and the AI reads the order number and product lines. Drivers photograph the signed order through a WhatsApp link, an admin approves it, and stock is deducted. Every day the system records stock levels, works out how many days each product will last and marks what needs reordering.
| Client | KKPL, interior fittings and furniture hardware, Singapore |
|---|---|
| What we built | One web system for sales, warehouse and admin teams, plus a driver upload page that needs no login |
| Automated | Lead creation and alerts, lead qualification and routing, document reading, stock snapshots, days-of-stock and reorder flags, incoming stock |
| AI used | Google Gemini: qualifying enquiries, summarising chats, reading delivery orders and other documents, and product search. Stock forecasting is a calculation, not AI. |
What the system had to handle
- WhatsApp enquiries reaching the right person quickly, with individuals and companies handled differently.
- Delivery orders and other paperwork read and entered without retyping every line.
- Proof of delivery collected from drivers and checked before stock changes.
- Clear visibility of stock levels and what to reorder, before items run out.
How a WhatsApp enquiry flows
- A customer messages KKPL. The system creates a lead and alerts the team that a new enquiry is waiting.
- The AI asks one question. Is this for an individual or a company, and which company? It asks at most three times, then hands over.
- The lead is routed. Individuals go to a sales agent and companies go to an admin. Chats that go quiet are handed over after 30 minutes.
- Admins get a summary. A WhatsApp alert with an AI summary of the conversation so far.
- Sales takes over. Agents track each lead's progress and close or reopen it.
How a delivery order flows into stock
- Staff upload the delivery order. The AI reads the order number and product lines, and the stock on it is reserved.
- The driver gets a link. A WhatsApp link or QR code opens an upload page, with no login.
- The driver sends proof. Photos of the signed order, several pages if needed, are read, combined into one PDF and checked against the order.
- An admin approves. Stock is deducted, never more than is still outstanding on the order.
- Mismatches are flagged. If stock would go below zero, it is logged for the team to reconcile.
What runs automatically
| When this happens | The system does this |
|---|---|
| A customer sends a WhatsApp message | Creates a lead and alerts the team about new unassigned enquiries |
| The AI has qualified the lead | Routes individuals to sales agents and companies to admins, with an AI summary |
| A chat goes quiet for 30 minutes | Hands it over to a person |
| Each day | Sends a count of unassigned leads and saves a stock-level snapshot for every product |
| Stock analysis runs | Works out average daily usage and days of stock left, and flags each product as out of stock, danger, overstock or healthy |
| A product will run out soon | Marks the reorder as urgent (under 7 days) or plan (under 14 days) |
| A delivery order or document is uploaded | Reads the lines with AI for staff to confirm |
| An incoming shipment arrives | Adds the goods to stock |
Roles for every team
- Admins: dashboard, leads, inventory, delivery orders, documents, upcoming stock, logs and settings.
- Sales agents: their leads, a WhatsApp inbox and stock levels.
- Warehouse: inventory, delivery orders, documents and upcoming stock.
- View-only users: inventory and logs.
- Drivers: a one-page upload link, nothing to install.
What stays with the team
- The sales conversation after handover, and closing each lead.
- Approving or rejecting each driver's proof of delivery.
- Confirming what the AI read from each document, line by line.
- Manual stock holds, count corrections and fixing negative balances.
More than delivery orders
The same AI document reading handles invoices, credit notes, memos and supplier packing lists, and staff can photograph a stock count to update quantities. Product search understands what people mean, not just exact product codes. Every reading is shown to staff to confirm before anything changes.
What we built
- WhatsApp lead capture
- AI lead qualification (individual or company)
- Automatic lead routing and handover
- AI chat summaries for admins
- Lead journey tracking
- AI delivery order reading
- Driver proof upload by link or QR code
- Delivery order approval with stock deduction
- Stock in and out tracking
- Daily stock snapshots
- Days-of-stock and reorder flags
- Incoming shipment handling
- AI reading of invoices, credit notes and packing lists
- Smart product search
- Roles for admin, sales, warehouse and view-only
- Excel import and export, PDF reports
Could this work for your business?
The same setup fits distributors, suppliers and trading companies that take enquiries on WhatsApp, move stock through delivery orders and need to know what to reorder. Implementation starts from S$3,000, plus a monthly fee based on usage, and a typical system is ready for testing in 2 to 4 weeks. Not sure where to start? Read what to automate first.
Related: all case studies, AI automation in Singapore, WhatsApp automation, workflow automation in Singapore, how a project runs.
Still retyping delivery orders?
Tell us how a normal day runs today. We will show you which steps can run by themselves.
Message us on WhatsApp