Automating a broken process just gets you broken faster.

Most companies do not have a software problem. They have a process that nobody has ever written down, and a piece of software bought on the assumption that it would supply the process.

It does not. Software supplies speed. If the process underneath is broken, speed is the last thing you want.

We see it every week. The tool is new, the invoice for it has been paid, and the work is moving exactly the way it moved before, only faster and with more places for it to go wrong. Here are three versions of the same mistake.

The quoting tool

A shop buys quoting software because quotes take too long to get out. The software is good. It produces a clean quote in a fraction of the time.

But nobody trusted the old quotes. The estimator was working from numbers in his head, adjusting for things that never made it onto paper, and the office was quietly re-checking his work before anything went out. The new tool took that untrusted process and made it faster. Now the bad quotes go out the same day.

The problem was never the speed of the quote. It was that the pricing lived in one person's memory. That is what needed writing down first.

The scheduling app

A service company puts in a scheduling app because the schedule is a mess. The app is fine. It has a calendar, drag and drop, reminders, the works.

The schedule still lives in the office manager's head. She knows which tech will not work with which customer, which jobs always run long, and which addresses are forty minutes from anything. None of that is in the app, because nobody asked her to put it there, and there was nowhere to put it if they had. So she keeps her own version on paper and updates the app when she has time.

Now the company has two schedules. The one on the screen is wrong, and the one that is right is still in one person's head.

The inbox tool

A growing company adds a tool that drafts email replies because the inbox is out of control. It works. Replies go out faster.

The replies are still wrong, because the person who used to answer was guessing at half of them. Nobody had ever written down what the company actually says when a customer asks about lead times, or warranty, or what happens if the part is back-ordered. The tool learned from the guesses. Now the guesses go out in minutes instead of hours, and with more confidence.

What the three have in common

In each case, the company bought speed for a process that had never been looked at. The tool did exactly what it was supposed to do. The process it was laid over had never been written down, so the tool automated whatever was actually happening, including the parts nobody wanted.

Write down one process this week. Not five. One.

That is the rule we follow before we build anything for anyone, and it is the rule we would give any owner for free. One process, followed from start to finish and written down plainly, will tell you more about where your company is losing time and money than any demo, any audit, and any tool.

It takes about an hour. Pick the process people complain about most. Take one real unit of work, yesterday's quote request or this morning's service call, and follow it the way it actually happened. For each step, note who did it, which tool they used, how long it took, and how long it sat waiting before it started. Circle every place information got retyped and every place someone had to stop and ask a question.

Most companies find that the waiting time is three to five times the working time. The work was never the problem. The hand-offs were.

Now, and only now, decide what to build. Sometimes it is a tool you already pay for that nobody turned on. Sometimes it is connecting two systems. Sometimes it is one small piece of software built for this exact process. And sometimes the honest answer is that the process is fine and the problem is somewhere else. All four are cheaper than buying first.

The full version of this, with the one-page worksheet, is our guide The One Process Rule. It is free, and it takes about ten minutes to read.