"Ramlila katrni 4 bag 50 Lajawab sella 3 bag 48 Rajesh Kirana Susuwahi." This message is complete to the rep who sent it. It contains the brand, variety, quantity, and rate. What it does not contain: the pack size, the current cycle price, whether the rate was agreed at the door or assumed from memory, whether the shop owner confirmed the order or the rep is writing from a preliminary conversation. The rep knows all of this. They wrote none of it.
What the message actually is
When a field rep sends an order message, they are not filling out a form. They are summarizing a sales interaction that already happened, in the shortest form possible, while walking to the next stop. The compression is not laziness. It is efficiency between two people who share context.
"Ramlila katrni" means something specific to this rep and this owner. They both know it is the Ram Leela brand, Katarni variety, 26 kg bags, billed per bag, at the rate agreed in the current cycle. The shared context fills all the gaps the message leaves empty.
This is exactly how human communication works: fragments that are complete to people who share the same frame, and opaque to anyone who does not. A message between two colleagues ("same as last time but minus the red ones") is a complete instruction to them. To anyone reading it cold, it is meaningless.
The problem is not that the rep wrote a fragment. The problem is that the shared context making the fragment readable has a shelf life.
When context expires
The rep sends the message at 3:47 pm. They are in the market. The conversation with the shop owner happened twenty minutes ago. Everything is fresh: the quantity, the rate, whether the owner hesitated, whether there was an amendment at the door.
By 6 pm, the rep has visited four more shops and is headed home. The Rajesh Kirana order is still accurate in their memory, but it is no longer the top frame.
By the time the owner reviews orders in the evening, the rep is unavailable. The message is read, the gaps are filled by assumption (the usual bag size, the standard rate), and an invoice is created. If the assumptions are right, the invoice is correct. If the assumptions are wrong, the invoice is wrong. And the rep, who would have caught the error instantly at 3:47 pm, is now unreachable.
Three weeks later, when Rajesh Kirana disputes the bag count or the rate, both the rep and the owner are reconstructing a conversation from memory. The message is the only written record, and it does not resolve the dispute. It is the evidence everyone is arguing about.
Three ways intent fails
The brand collapses. Reps spell brand names phonetically, from memory, under time pressure. "Ramlila", "Ram Leela", "Ramlila katrni", "ramleela": all the same product in a rep's head. In any system treating them as text strings, they are four separate brands. The catalogue splinters. Analytics lie. The rep who sold the most of this brand disappears from the leaderboard because half their orders are filed under a spelling variant.
The rate drifts. Cycle rates update, but not everyone's mental rate sheet updates at the same time. A rep who has run this route for years knows the rice rate as a number they heard months ago. When the cycle changes, they may not update their default. The message says "@50" because that is the number in their head. The correct current rate is 48. Two rupees per bag, across fifty bags per day, across twenty working days: that is two thousand rupees of margin leakage per rep per month. It shows up nowhere until someone pulls a rate-versus-catalogue comparison and notices the drift.
The intent is optimistic. This is the hardest one. A rep who visited a shop, had a good conversation, and left with what felt like a soft commitment sometimes writes the order as if the commitment were confirmed. The message is not dishonest. It is the rep's genuine read of the interaction. But the shop owner's read was different: they were interested, not committed. The delivery arrives, the shop owner says they never agreed to that quantity, and now everyone is having a conversation that should have happened at the door, not at the loading dock.
The message is intent. It carries what the rep meant to communicate, not a verified record of what was agreed.
The correction window
There is one moment when all three failure modes are correctable: immediately after the rep sends the message.
If the system reads the message and sends back a structured interpretation within seconds, the rep is still in the area. They still remember the conversation. Correcting a wrong product or a wrong rate takes ten seconds. The context is live.
That window closes fast. After the rep reaches the next shop, the previous order recedes. After the end of the day, it is a memory competing with thirty other memories. After three weeks, it is a disputed invoice.
The correction has to happen at the source: not when the invoice is printed, not when the shop owner calls, but at the moment the rep hits send, while the context is still in the room. A structured confirmation in seconds is not a verification burden. It is the ten-second check that prevents a three-hour dispute.
Distribution businesses that close the intent gap at the source see two things happen. Billing disputes go down because the record reflects what was actually agreed. And margin leakage from rate drift becomes visible for the first time, because the system is now comparing what the rep wrote against what the catalogue says the rate should be, on every single order.