Should I Automate This, or Fix It First?

On calls I hear the same request again and again: "My team is drowning. Automate everything." I understand it. But my first answer is almost always a question: "What exactly do you want to repeat?"

The short answer is this: automation repeats a process exactly as it is. It does not fix it.

The GPS with the wrong address

Picture typing the wrong address into your GPS. It doesn't slow you down. On the contrary, it takes you to the wrong place faster, calmly and with total confidence, and it even tells you when to turn.

That is what automation does to a broken process. You get a faster mistake. That is why you check the address before you press "Go." For us, the address is the analysis of the process.

Three questions before I say yes

1. Is the process written down, and does everyone do it the same way? If two people do the same job in two different ways, the machine has nothing to copy. First you decide how it is done. Say orders are forwarded from an inbox into a spreadsheet. One colleague puts the phone number in one column, another writes it in the notes, a third forgets it. Put automation on top of that and it will carry the same mess along, only faster and without anyone noticing.

2. How often does it happen, and how long does it take? Multiply how many times a month it happens by how many minutes each time. Say 20 times at 6 minutes is 2 hours a month. Even xkcd (a popular web comic) has a chart for this: a task you do once a month that saves you 5 minutes is worth at most 5 hours of work over five years. Rare and small things rarely pay back.

3. Does it need a person's judgement? If every case is decided by the situation, that part stays with a person.

These three questions are not a formality: they are where you see whether the process has to be put in order first.

There is a second way to look at all this: I sort processes by value and difficulty. High value and easy: do it now. High value and hard: plan it with preparation. Low value and easy: not a priority. Low value and hard: usually leave it as it is.

What we see in our work

These are only some of the cases from our work, not the whole picture.

• A real estate agency. Clients came from many ads, and some of them were getting lost. Our first step was not automation. It was putting the data in order in Google Sheets (Google's online spreadsheet). Then we moved it to Airtable (an online database). Instead of building more and more automations to patch the problem, we changed the structure. The rule I took from it: automate mechanical tasks, not sales. The full case is in the agency article.
• A freight company. In the system that receives customer requests, a ticket containing words like "urgent," "complaint" or "legal" goes straight to a person, by design. The rollout was careful too: first only the safest categories, everything else stayed manual. How the safeguards are built is in the freight company case.
• Sometimes the wrong step is AI where plain code is enough. That is a different story, told in "When AI Agents Cost More Than They're Worth".

In all three, the work did not start with the question "what should we automate?" It started with "what is actually happening?" The answer to the second question often changes the answer to the first. In one case it turned out the structure had to change, in another that part of the work had to stay with people.

Your real alternatives

Automation is not your only choice. Here are the others, and honestly, sometimes they win.

• Do nothing. If it hurts rarely and costs little, that is perfectly reasonable.
• A paper checklist. Five to ten steps on one page, pinned next to the computer. Often that is half the solution.
• One responsible person. Someone who knows the process is theirs and keeps an eye on it.
• A feature your software already has. For example rules (filters) in Gmail or a form that fills the spreadsheet by itself. Often it is already in the software you pay for.
• Automation. When the process is clear, frequent and needs no judgement. We build it in n8n (a tool that connects apps and runs automatic workflows).

I am not telling you this to talk you out of automation. I say it because the most expensive decision is the one you buy before you have seen what exactly you are buying. If one of these options is enough, it is better to hear it from me before you spend money. More on when automation is not worth it is in "What Is Business Process Automation, and Where Do I Start?".

Why we start with analysis

That is why we always work in the same order. First, analysis of the process: how it is done today, where time is lost and where mistakes happen. Only after that do we automate the whole process, not separate pieces. That way you know what you are buying before you pay for it.

My analysis starts with the Process X-ray, and how we document it is in "What is PDD and Why It's Critical".

What to do tomorrow

1. Write one process on one page, in 5 to 10 steps. Ask two more people to write the same one without seeing yours. Compare the three pages. The differences are your real problem.
2. Put one number on one process. How many times a month it happens, multiplied by the minutes each time. That gives you hours per month.
3. Ask two questions. Ask the software you use: "Can you already do this?" Ask anyone selling you automation: "Which part would you leave manual, and why?"

If you are not sure where to begin, start here. If it is not worth it, you will hear that on the call.