Home/Blog/Automation

The two systems that didn't talk

They had one system for sales and another for invoicing. The problem was the dead space between them: a person retyping by hand what a machine could move on its own.

At Nuria's company, which sold spare parts for agricultural machinery in Albacete, central Spain, there were two systems. One for sales, where the reps entered the orders. Another for invoicing, where admin produced the invoices. Both were good. Each did its own job well. The problem wasn't in either of them. It was in the gap between them.

The gap was filled by Loli. Every morning, Loli opened the sales system, looked at the previous day's orders, and copied them across, one by one, into the invoicing system. Customer, address, products, quantities, prices. She typed it all again. The same thing that was already written in one place, she wrote again in another. Around forty orders a day. Two hours every morning.

—It's just that the systems don't talk —Loli would say, like someone explaining that it rains.

And she was right. They didn't talk. They were from two different makers, bought in different years, and neither knew the other existed. Between them was a person acting as a bridge. A bridge of flesh and blood who got it wrong now and then, the way we all do when we copy numbers by hand forty times in a row.

What the bridge cost

Nobody at the company had added up what that bridge cost, because it was a cost that showed up on no invoice. But it was there.

There were Loli's two hours every morning, of course. Ten hours a week. One person, a quarter of their working time, spent copying from one place to another what was already written.

There were the errors. Not many, but just enough to do damage. An extra zero in a quantity. A price from the previous order that stuck. A delivery address that didn't get updated. Each of those errors was discovered late, when the customer called annoyed, and fixing it cost more than the two hours of copying: it cost the call, the apology, the credit note, and a little trust.

And there was the delay. Because Loli moved the orders in the morning, an order closed at five in the afternoon wasn't invoiced until the next day. And if Loli had the day off, until Monday. The company collected its money later than it sold, not because the customer was slow to pay, but because the invoice was slow to go out.

What Nuria asked for

Nuria called me about something else. She wanted to know if I'd build her a new system that did both things, sales and invoicing, all together, so she could get rid of the two old ones. She was ready to throw out what she had and start from scratch.

I told her not to do that. That her two systems worked well, that her people knew them, that replacing them would be months of upheaval and training to solve a problem that wasn't in the systems. The problem was the bridge. And you don't fix a bridge by demolishing both banks. You fix it by building the bridge.

—And that can be done?
—That's the easiest part of everything you've told me.

Building the bridge

What we built wasn't a new system. It was an automation: a small piece that sits between the two systems and does exactly what Loli did, but without typing and without mistakes.

It works like this. When a rep closes an order in the sales system, the automation knows at once. It takes the order's data — the customer, the products, the quantities, the prices — and enters it into the invoicing system, in the right place, without anyone copying anything. The invoice is ready to review in seconds, not the next morning.

Most systems today, even when they don't talk to each other out of the box, have a service door through which information can be passed in and out in an orderly way. The maker doesn't have to do anything. That door is already there. What you build is the messenger that goes from one door to the other, carrying the data, checking it arrives properly, and flagging anything that doesn't add up.

Because we added that too: if an order brought something odd — a customer that didn't exist in invoicing, a product with no price — the automation didn't invent it, nor did it let it through. It set it aside and told Loli. Just as Loli used to stop when something didn't add up, now the machine stopped and told her.

What happened to Loli

The question that always comes up, and that's best not dodged: what about Loli?

Loli wasn't surplus. Loli knew the customers, the special prices, which order was urgent and which could wait, which customer paid well and which had to be watched. She had all of that in her head and didn't use it, because her morning went on copying numbers.

When the automation took over the copying, Loli took over the rest. Reviewing the orders the machine set aside, which were the interesting ones, the ones with something odd. Calling the customers with overdue invoices. Balancing the month's accounts without waiting for the last day. She stopped being the person who copied and became the person who kept control. Same salary, far better used.

Four lines in a notebook

A month later, Nuria did a sum in a notebook, the kind you do for yourself. Loli's ten hours a week, recovered. The copying errors, gone. The invoices, going out the same day as the sale, which brought collections forward by an average of a day and a half across everything they invoiced. And a change she didn't know how to write down: Loli was happier.

She told me over the phone, and at the end she said something I liked.

—I thought I needed a new system. What I needed was for the two I have to say good morning to each other.

I told her that was, word for word, the best definition of integration I'd ever heard. That most companies don't need more software. They need the software they already have to stop ignoring each other. The work isn't buying another tool. It's taking the person out of the gap, and letting the data cross on its own.

Before you buy another system

When something isn't working, the temptation is to buy a new tool. But often the problem isn't in the tools: it's in the gap between them, that dead space where a person copies by hand what's already written somewhere else.

Those gaps can almost always be closed with an automation: a small piece that moves the data from one system to another on its own, checks it arrives properly, and flags anything that doesn't add up. Cheaper than a new system, with no throwing out what already works and no months of training. Before you switch tools, check whether what you're missing is the bridge.

Does someone in your company copy data from one system to another?

If there's a person spending hours moving information from one system to another by hand, that bridge can almost always be automated. We can look at it together.

Let's talk