How to Stop Retyping Orders Between Systems

An operations lead at a licensed apparel company told me where her sales team's booking window was going. Not into selling. Into entering what had already been sold.

I asked how many people were doing manual data entry and order processing. She said "double digits."

Then she said the thing that made the whole engagement obvious. About eighty-five percent of their orders follow a predictable pattern. Same shapes, same sizes, same handful of decisions. And every one of them was being retyped by a person, because the order started in one system and had to end up in another, and the two would not talk.

That is not an automation problem in the way people usually mean it. Nobody needed a model to decide anything, because the decisions were already made. What was missing was translation.

Translation is the most expensive work nobody has costed

Here is the shape of it. Once you see it, you will find it in your own company by Thursday.

Information is created in system A. It is needed in system B. The two have different ideas about what a record is, what a required field is, and what a date looks like. So a person opens both, reads one, and types into the other. They add no judgment. They are a protocol converter made of human being.

At that apparel company the journey ran from an online store and a CRM to the operations platform the facility actually runs on. Every hop needed manual handling to get the data into the shape the next system would accept.

Nobody in the building called this a problem, because nobody could see all of it at once. Each person only saw their own hop. The sales coordinator saw a spreadsheet. The person downstream saw a queue. The cost only appears when you add the hops up.

Why it survives so long

Translation work is invisible to the three places a company usually looks.

It does not show up in headcount planning, because nobody was hired to do it. It accumulated. A person hired to manage accounts is now spending a third of their week retyping.

It does not show up in software spend, because the fix is not a product. Both systems are already paid for. What is missing is the thing between them, and no vendor sells you that.

And it does not show up in the complaint channel, because the people doing it have stopped noticing. When I ask "how long does that take you," the answer is usually a shrug and then a number that surprises the person saying it.

That is why discovery is interviews, and why the interviews are with the people doing the work rather than the people who own the budget. Nobody is hiding this. No single person has ever had cause to add it up.

What we build first, and why

The instinct is to start with the hardest case: the fifteen percent, the odd orders, the exceptions, the ones that need a human to think.

That is the wrong end. Start with the eighty-five percent.

Standard work is where automation is safe, because the pattern is known and the failure mode is visible. If a standard order comes through wrong, everyone downstream notices immediately. The exceptions are exactly where an automated system will fail quietly, which is the only kind of failure that really costs you.

So the first build takes the predictable path end to end. Order lands, gets shaped, arrives in the operations platform in the format it expects, with a record of what happened. The fifteen percent still goes to a person, and now that person has time to actually look at it, which is what you wanted from them in the first place.

The second thing that happens is better than the first. Once the standard path is automated, the exceptions stop hiding inside the volume. You can finally count them. Usually a third of them turn out to be one recurring problem with a customer setup, and that is a twenty-minute fix that nobody could see before.

What to do this week

Pick one flow where information moves between two systems. Just one.

Sit with the person who moves it and time a single pass. Not their estimate. Time it, once, with a stopwatch, on a real record.

Then multiply by volume. That is the number. In my experience it is between three and ten times what everyone assumed, and the person who does the job every day is as surprised as their executive.

You do not need an AI strategy to have that conversation. You need forty-five minutes and the willingness to hear an uncomfortable number.

The strategy comes after, and it is usually shorter than people expect. Automate the predictable path. Route the exceptions to a person with a name. Measure the queue.