Picture a food supplies distributor. Thirty staff, a few hundred restaurant and café customers. Orders arrive every way imaginable: chat messages, voice notes, photos of handwritten lists, phone calls. Every one gets retyped by hand. Late-night orders get missed. Wrong items go out. The phone rings all morning with "where's my delivery?". Payments come in late. And the order desk is under constant pressure.
The steps, and why they matter
Listen to how orders really arrive (Fix the right problems first)
Teach the AI the catalogue and every customer (AI that knows every customer)
Orders in, in any form (No more retyping)
Every call answered (Every order captured)
Deliveries customers can see (Customers kept informed)
Payments without the chasing (Faster payments)
Customers who keep ordering (More repeat orders)
An order desk people want to work at (Staff who stay).
Where people stay in control
Messages become draft orders. Anything uncertain goes to a person to check. Invoices follow delivered quantities, and account issues receive human review.
Start with one outlet, clinic, team or delivery route. Train the team, measure and adjust before extending the approach. The video’s end state illustrates a direction, rather than reporting achieved results.
Read the full video transcript
At eleven forty at night, a restaurant sends a voice note: two cases of cooking oil, five kilos of chicken, and "the usual" sauces. Tomorrow morning, someone has to work out what "the usual" means. Here's how AI can turn a distributor around, step by step. Picture a food supplies distributor. Thirty staff, a few hundred restaurant and café customers. Orders arrive every way imaginable: chat messages, voice notes, photos of handwritten lists, phone calls. Every one gets retyped by hand. Late-night orders get missed. Wrong items go out. The phone rings all morning with "where's my delivery?". Payments come in late. And the order desk is under constant pressure. It starts by sitting with the order desk: counting how orders arrive, how long each one takes to key in, and how many need correcting later. Then the awkward cases: "the usual", an item that's out of stock, an urgent same-day delivery, a new customer with no account, an account over its credit limit. Each step gets sorted: automate it, let AI do it with a person checking, or keep it human. Then the AI learns what an experienced order clerk knows: every product and pack size, each customer's usual order and delivery days, which items can be swapped when stock runs out, the cut-off times, the credit terms. Now "the usual" actually means something. Here's the heart of it. Whether an order comes as a message, a voice note in Malay, English or Mandarin, or a photo of a handwritten list, the AI turns it into a draft order: products matched, quantities checked, "the usual" filled in from history. And here's the expert part: it shows how sure it is. Anything it isn't sure about goes to a person to check. Then the order goes straight into the distributor's existing system, and the customer gets a confirmation. Calls get answered too, at any hour. "Do you have stock?" "Can I add one more case?" "What time is my delivery?" The AI answers from live stock and delivery information. Price disputes and account questions go to a person. Most "where's my delivery?" calls disappear when customers simply know. So each customer gets a message when their delivery is on the way, with an arrival window. Drivers capture proof of delivery. If an item was short, it's flagged straight away, not discovered weeks later in a dispute. Invoices are built from what was actually delivered, so they're right the first time. Customers get a friendly reminder before a payment is due, by message and email. And if an account goes over its limit, a person reviews it before the next order. No harsh automation, just no more forgotten invoices. Then the relationship. The system notices patterns, so a customer who orders oil every Monday gets a timely nudge: "the usual?". Customers who've opted in hear about new products that suit their kitchen. And new restaurants find the distributor too, through an online catalogue, a proper business profile, and ads aimed at new restaurants opening in its delivery areas. And the team? Order clerks stop retyping and start handling the exceptions and the relationships. Sales reps visit customers instead of chasing orders. The warehouse gets clean pick lists earlier. When the pressure comes off, good people stay. One delivery route goes first. Everything is measured and adjusted before the rest follow. So here's the end of the turnaround. Orders in within minutes, in any form. Every order captured, even at midnight. Fewer mistakes. Customers who know where their delivery is. Faster payments. And a team that wants to stay. That's what AI can do for a distributor. It doesn't replace the people who know your customers. It takes away the retyping and the chasing, so they can look after them. If you're wondering what this could look like in your business, I'm always happy to talk it through on LinkedIn.
